Upload
phungkhanh
View
216
Download
0
Embed Size (px)
Citation preview
Embrandiri, Manoj (2014) Implementation and in-depth analyses of a battery-supercapacitor powered electric vehicle (E-Kancil). PhD thesis, University of Nottingham.
Access from the University of Nottingham repository: http://eprints.nottingham.ac.uk/14269/1/PhD_thesis_full_%28Manoj_Embrandiri_June_2014%29.pdf
Copyright and reuse:
The Nottingham ePrints service makes this work by researchers of the University of Nottingham available open access under the following conditions.
· Copyright and all moral rights to the version of the paper presented here belong to
the individual author(s) and/or other copyright owners.
· To the extent reasonable and practicable the material made available in Nottingham
ePrints has been checked for eligibility before being made available.
· Copies of full items can be used for personal research or study, educational, or not-
for-profit purposes without prior permission or charge provided that the authors, title and full bibliographic details are credited, a hyperlink and/or URL is given for the original metadata page and the content is not changed in any way.
· Quotations or similar reproductions must be sufficiently acknowledged.
Please see our full end user licence at: http://eprints.nottingham.ac.uk/end_user_agreement.pdf
A note on versions:
The version presented here may differ from the published version or from the version of record. If you wish to cite this item you are advised to consult the publisher’s version. Please see the repository url above for details on accessing the published version and note that access may require a subscription.
For more information, please contact [email protected]
i
IMPLEMENTATION AND IN-DEPTH ANALYSES OF
A BATTERY-SUPERCAPACITOR POWERED
ELECTRIC VEHICLE (E-KANCIL)
Manoj Embrandiri
Thesis submitted to the University of Nottingham for
the Degree of Doctor of Philosophy
December 2013
ii
ABSTRACT
This thesis contributes to the research issue pertaining to the management of
multiple energy sources on-board a pure electric vehicle; particularly the energy
dense traction battery and the power dense supercapacitor or ultracapacitor. This is
achieved by analysing real world drive data on the interaction between lead acid
battery pack and supercapacitor module connected in parallel while trying to fulfil
the load demands of the vehicle.
The initial findings and performance of a prototype electric vehicle conversion of a
famous Malaysian city car; the perodual kancil, is presented in this thesis. The 660
cc compact city car engine was replaced with a brushless DC motor rated at 8KW
continuous and 20KW peak. The battery pack consists of eight T105 Trojan 6V, 225
Ah deep cycle lead acid battery which builds up a voltage of 48V. In addition to this,
a supercapacitor module (165F, 48V) is connected in parallel using high power
contactors in order to investigate the increase in performance criteria such as
acceleration, range, battery life etc. which have been proven in various literatures
via simulation studies. A data acquisition system is setup in order to collect real
world driving data from the electric vehicle on the fly along a fixed route. Analysis of
collected driving data is done using MATLAB software and comparison of
performance of the electric vehicle with and without supercapacitor module is
made.
Results show that with a parallel connection, battery life and health is enhanced by
reduction in peak currents of up to 49%. Peak power capabilities of the entire
iii
hybrid source increased from 9.5KW to 12.5KW. A 41% increase in range per charge
was recorded. The author of this work hopes that by capitalizing on the natural
peak power buffering capabilities of the supercapacitor, a cost effective energy
management system can be designed in order to utilize more than 23.6% of the
supercapacitor energy.
iv
PUBLICATIONS
During the course of this research work, the following papers were accepted for
publication in international conferences and journals:
1. M;ミラテくEが Dキミラ Iゲ;が RラゲWノキミ; AヴWノエキが さ“┌ヮWヴI;ヮ;IキデラヴっB;デデWヴ┞ エ┞HヴキS Pラ┘WヴWS
EノWIデヴキI BキI┞IノW ┗キ; ; “マ;ヴデ Bララゲデ Cラミ┗WヴデWヴざ ア EV“-25 Shenzhen, China,
Nov. 5-9, 2010 The 25th World Battery, Hybrid and Fuel Cell Electric Vehicle
Symposium & Exhibition.
2. M;ミラテくEが Dキミラ Iゲ;が RラゲWノキミ; AヴWノエキが さPWヴaラヴマ;ミIW ラa ;ミ EノWIデヴキI VWエキIノW
Cラミ┗Wヴゲキラミ ;デ デエW Uミキ┗Wヴゲキデ┞ ラa Nラデデキミェエ;マが M;ノ;┞ゲキ;ざ ヮヴWゲWミデWS ;デ ┘ラヴノS
conference of fuel economy and sustainable road transport, Institution of
Mechanical Engineers (IMechE), Pune India.
3. Embrandiri, Manoj, Dino Isa, and Roselina Arehli. "An Electric Vehicle
Conversion using Batteries and Ultracapacitors." Journal of Asian Electric
Vehicles 9.2 (2011): 1521-1527.
4. Manoj.E, Dino Isa, RラゲWノキミ; AヴWノエキが さ“┌ヮWヴI;ヮ;IキデラヴっB;デデWヴ┞ エ┞HヴキS Pラ┘WヴWS
EノWIデヴキI BキI┞IノW ┗キ; ; “マ;ヴデ Bララゲデ Cラミ┗WヴデWヴざ WラヴノS EノWIデヴキI VWエキIノW Jラ┌ヴミ;ノ
Vol. 4 - ISSN 2032-6653 - © 2010 WEVA page 280.
v
ACKNOWLEDGEMENTS
I would like to express my sincere gratitude to Prof. Dino Isa, firstly for giving me
this opportunity and secondly for his constant support, guidance and friendship
throughout the course of this research work. It has come a long way.
Special thanks to my dad, mom and sisters for their much needed moral and
financial support over the years. Thank you for believing in me. To the Indrans,
thank you for being my family in Malaysia. The amount of support I have received
from you guys over the years is overwhelming and definitely not quantifiable. To my
fellow PhD colleagues Ahmida and Zhen Wei, we have had some good laughs over
the years. I wish you all the best in the future.
The equipment that was used to carry out this research work was funded by a
research grant e-Science from the ministry of science Malaysia (MOSTI) and the
facilities at the University of Nottingham Malaysia campus. Several aspects of this
project would not have been achieved without inputs from people like Mr. K.
Parthipan and Mr. Joseph Koh etc.
Finally and most importantly, you, yes you; Shanu Indran, the love of my life, thank
you for tolerating me and for being there through thick and thin.
vi
TABLE OF CONTENTS
ABSTRACT ii
PUBLICATIONS iv
ACKNOWLEDGEMENTS v
TABLE OF CONTENTS vi
LIST OF FIGURES ix
LIST OF TABLES xiv
LIST OF ACRONYMS xiv
CHAPTER 1 INTRODUCTION 1
1.1 Introduction to the Thesis 1
1.2 Research Issues Addressed by this Work 3
1.3 Research Objectives 5
1.4 Research Scope 6
1.5 Thesis Contribution 7
1.6 Thesis Organisation 8
vii
CHAPTER 2 BACKGROUND AND LITERATURE REVIEW 10
2.1 Background 10
2.1.1 Electric Vehicles 11
2.1.2 Hybrid Electric Vehicle Drive Configurations 16
2.1.3 Battery Technologies for Electric Vehicles 19
2.1.4 Supercapacitor/Ultracapacitor 27
2.2 State-of-the-Art on Hybrid/Multiple Energy Sources 35
for Electric Vehicles
CHAPTER 3 METHODOLOGY 80
3.1 Electric Vehicle Design Criteria 82
3.2 Software Modelling of a Small Electric Vehicle 87
3.3 Integrating Supercapacitors into an Electric Vehicle 91
(Hardware Description)
3.3.1 Experimental Test Vehicle 95
3.3.2 Electric Motor Coupling and Mounting 99
3.3.3 Electrical Wiring 102
3.3.4 Installation and Testing of the Supercapacitor Module 108
viii
3.3.5 Brake System Modification 110
3.4 Data Acquisition and Testing 112
3.4.1 Driving Data Collection 117
CHAPTER 4 RESULTS AND DISCUSSION 122
4.1 Battery-Supercapacitor Implementation 123
4.1.1 Results: Battery Only (Supercapacitor OFF) 132
4.1.2 Results: Battery + Supercapacitor in Parallel 142
4.1.3 Results: Battery + Supercapacitor in Parallel (In a Mixed Cycle) 155
4.1.4 Results: Range Test 163
4.2 UNMC Drive Cycle Analysis 165
CHAPTER 5 CONCLUSION AND FUTURE WORK 178
5.1 Conclusion 178
5.2 Future Work 183
REFERENCES 184
APPENDIX 194
ix
LIST OF FIGURES
CHAPTER 2
Fig. 1 The Basic Electric Vehicle Configuration 12
Fig. 2 Emissions: EV vs. ICE 14
Fig. 3 Series Hybrid Drive 16
Fig. 4 Parallel Hybrid Drive 17
Fig. 5 A Comparison of Energy densities of various batteries 27
Fig. 6 Electric Double Layer Capacitor Structure 28
Fig. 7 Electric Double Layer Modeled as Series of Parallel RC Circuits 31
Fig. 8 1st Order Model of the Supercapacitor 31
Fig. 9 Non Ideal (transmission line) Model of the Supercapacitor 32
Fig. 10 A Ragone Chart 33
Fig. 11 General Overview of an Energy Management System 36
Fig. 12 Power arbitration concept for multiple energy sources 37
Fig. 13 Supercapacitor-battery parallel circuit and its Thevenin equivalent 39
in frequency domain
Fig. 14 Battery + Supercapacitor Series Configuration 42
Fig. 15 Direct Parallel Connection of Battery bank (b) with Supercapacitor 43
Bank (c) supplying a load.
Fig. 16 Passive (left) and Active (right) Configurations 43
Fig. 17 Passive (left) and Active (right) Current waveforms 45
Fig. 18 Constant Current Regulator based Configuration. 48
Fig. 19 Block Diagram and Control Scheme of ZEBRA + Supercapacitor 49
by Dixon et al.
x
Fig. 20 Flow chart of Control developed for battery supercapacitor 58
Shelby cobra
Fig. 21 14 features selected for roadway prediction 59
Fig. 22 UDDS cycle segmented by a window size and a time step. 61
Fig. 23 Driver Style (DS) Classification Algorithm Proposed by Y.L. Murphey 63
Fig. 24 Architecture of the Intelligent Energy Management Agent (IEMA). 65
Fig. 25 Block diagram of Fuzzy Logic Supervisory Controller by A.A Ferreira 66
Fig. 26 Fuzzy Membership Functions and Rule Base by A.A Ferreira 68
Fig. 27 Block diagram of driving load forecasting of HEV. 72
Fig. 28 Forecasting energy usage in HEV based on 3D terrain maps 75
Fig. 29 Drive cycle analysis; fuzzy classification of drive pulses (DP). 79
Fig. 30 A mind map of the literature review conducted 81
CHAPTER 3
Fig. 33 The forces acting on a car 81
Fig. 34 Simulating the Power (KW) needed by a Small Vehicle 88
for low Speed Drive Cycle.
Fig. 35 Simulating the Depth of Discharge of a Lead acid 89
Battery pack versus distance travelled (range)
Fig. 36 Simulating the Current drawn from the battery pack @ 48V 207Ah 90
Fig. 37 Setup for a Sample Electric Vehicle laboratory Test Bench 92
Fig. 38 Work flow for EV conversion 94
xi
Fig. 39 Estimated Discharge Curve for Trojan T105 deep cycle battery 97
Fig. 40 Estimated Life Cycle Curve for T105 battery 97
Fig. 41 Mechanical Part of EV Conversion 98
Fig. 42 Electric Motor with coupler (center), custom made
adaptor plate (top right) and a spacer (bottom left) 100
Fig. 43 Brushless DC Motor successfully mounted, adaptor plate matches
the bolt pattern of the original transmission block. Reinforcement
bracket is added to make the mount more stable. 100
Fig. 44 500A Motor Controller successfully mounted in the same position
as the radiator of a conventional vehicle. 101
Fig. 45 Complete Wiring Diagram for Electric Vehicle Conversion 107
Fig. 46 BMOD0165P048; 165F, 48.6V supercapacitor which weighs 15kg 108
Fig. 47 BMOD0165P048 installed just above the electric motor.
Top right is the high current contactor which is activated
by a 12V coil and a manual switch on the dashboard of the vehicle 108
Fig. 48 Constant Current Discharge Profile for BMOD0165P048
Supercapacitor Module 109
Fig. 49 Vacuum pump wiring diagram for brake system modification 110
Fig. 50 Electric Vacuum Pump (left) connected to the vacuum reservoir
on one side and the brake booster on the other. A pressure sensor
turns the pump on and off depending on the set level. 110
Fig. 51 Front view of the EV with all the necessary connections in place. 111
Fig. 52 DataTaker DT82E Data logger with flash USB storage logs
battery pack current and voltage, supercapacitor current
and voltage, GPS timestamps as well as GPS speed. 112
xii
Fig. 53 Hall-effect current transducers used to monitor battery
and supercapacitor currents. 113
Fig. 54 Differential Wiring diagram for Current and Voltage
Transducer to data logging device 114
Fig. 55 Dashboard Mounted Display 114
Fig. 56 Wiring diagram for GPS antenna and MAX232 level converter 114
Fig. 57 The UDDS drive cycle and some of its features 115
Fig. 58 The HWFET drive cycle and some of its features 118
Fig. 59 Fixed Driving route for EV testing at the University 118
Fig. 60 Testing and Data Collection With and Without Supercapacitor 120
Fig. 61 Snapshot of the data obtained from the logging system at 0.33Hz 120
CHAPTER 4
Fig. 62 UNMC Drive Cycle Trip 1 124
Fig. 63 UNMC Drive Cycle Trip 2 125
Fig. 64 UNMC Drive Cycle Trip 3 126
Fig. 65 UNMC Drive Cycle Trip 4 127
Fig. 66 UNMC Drive Cycle Trip 5 128
Fig. 67 Frequency distribution (histogram) of UNMC drive cycle TRIPS 1-5. 129
Fig. 68 Battery Voltage 134
Fig. 69 Frequency Distribution of Battery Voltage 135
Fig. 70 Battery Current 136
Fig. 71 Frequency Distribution of Battery Current 137
Fig. 72 Instantaneous Power 138
Fig. 73 Frequency Distribution of Instantaneous Power 139
Fig. 74 Vehicle Acceleration 140
xiii
Fig. 75 Frequency Distribution of Vehicle Acceleration 141
Fig. 76 Battery + Supercapacitor Voltage 145
Fig. 77 Frequency Distribution of Battery + Supercapacitor Voltage 146
Fig. 78 Battery + Supercapacitor Current 147
Fig. 79 Frequency Distribution of Battery + Supercapacitor Current 148
Fig. 80 Instantaneous Power delivered by Battery + Supercapacitor Hybrid 149
Fig. 81 Frequency Distribution of Battery + Supercapacitor Instantaneous
Power 150
Fig. 82 Effective Instantaneous Power delivered by
Battery + Supercapacitor Hybrid 151
Fig. 83 Frequency Distribution of Effective Instantaneous Power 152
Fig. 84 Vehicle Acceleration 153
Fig. 85 Frequency Distribution of Vehicle Acceleration 154
Fig. 86 Battery + Supercapacitor Voltage (Mixed Cycle) 158
Fig. 87 Battery + Supercapacitor Current (Mixed Cycle) 159
Fig. 88 Battery + Supercapacitor Instantaneous Power (Mixed Cycle) 160
Fig. 89 Effective Instantaneous Power of Hybrid Source (Mixed Cycle). 161
Fig. 90 Vehicle Acceleration (Mixed cycle) 162
Fig. 91 Dividing the UNMC Drive Cycle into Microtrips 166
Fig 92: Stem Plot of Max. Speed & Average Speed versus 52 Microtrips 172
Fig 93: Stem Plot of Max. & Average Acceleration versus 52 Microtrips 173
Fig 94: Stem Plot of Max. & Average Jerk versus 52 Microtrips 174
Fig 95: Stem Plot of Max. & Average Power versus 52 Microtrips 175
Fig 96: Comparing the Acceleration and Jerk Plots with the Instantaneous
Power by Supercap & Battery. 177
xiv
LIST OF TABLES
Table 1 : Specific battery parameters in comparison with supercapacitor 34
Table 2 : Comparing the two control strategies developed by Carter et al 50
Table 3 : Common types of kernel functions 85
Table 4 : Technical Specifications of the Perodua Kancil 96
Table 5 : EV conversion parts and specifications 105
Table 6: Summary of the main characteristics of the
UNMC drive cycle for 5 different trips 130
Table 7 : Summary of results (battery only) 132
Table 8 : Frequency table of Battery Voltage 135
Table 9 : Frequency table of Battery Current 137
Table 10: Frequency table of Instantaneous Power delivered by Battery 139
Table 11: Frequency table of Vehicle Acceleration 141
Table 12: Summary of results (battery + supercapacitor) 142
Table 13: Frequency table of Battery Voltage 146
Table 14: Frequency table of Supercapacitor Voltage 146
Table 15: Frequency table of Battery Current 148
Table 16: Frequency table of Supercapacitor Current 148
Table 17: Frequency table of Power delivered by Battery 150
Table 18: Frequency table of Power delivered by Supercapacitor 150
Table 19: Frequency table of Effective Power delivered by Hybrid Source 152
Table 20: Frequency table of Vehicle Acceleration 154
Table 21: Summary of results (Mixed cycle I) 156
Table 22: Summary of results (Mixed Cycle II) 157
Table 23: Summary of the Range Test for electric kancil 163
xv
Table 24: Statistical features of UNMC drive cycle: Microtrips 1:10 167
Table 25: Statistical features of UNMC drive cycle: Microtrips 11:20 168
Table 26: Statistical features of UNMC drive cycle: Microtrips 21:30 169
Table 27: Statistical features of UNMC drive cycle: Microtrips 31:40 170
Table 28: Statistical features of UNMC drive cycle: Microtrips 41:52 171
Table 29: Characteristics of the UNMC drive cycle 185
1
CHAPTER 1: INTRODUCTION
1.1 INTRODUCTION TO THE THESIS
This thesis is dedicated to the research, modelling and eventual implementation of
a battery-supercapacitor hybrid energy source in order to power an electric vehicle.
This is in recognition of the assertion that the electric vehicle will be a significant
stakeholder in future transportation systems.
A ェヴラ┘キミェ IラミIWヴミ キミ デラS;┞げゲ ┘ラヴノS キゲ Wミ┗キヴラミマWミデ;ノ ヮヴラデWIデキラミ ;ミS WミWヴェ┞
conservation. Automotive manufacturers are developing alternatives to existing
fossil fuel-driven vehicles. This has paved way for the development of Electric
Vehicles (EV) and Hybrid Electric Vehicles (HEV). While HEVs tend to reduce the
emissions from internal combustion vehicles as a result of greater fuel efficiency,
they do not completely solve the problem. Electric vehicles on the other hand are
much more energy efficient, produce absolutely no tail pipe emissions and requires
less maintenance as compared to the conventional internal combustion engine (ICE)
vehicles [2]. However, the reason the automotive industry has not gone pure
electric or able to compete favourably with existing gasoline cars, lies in the
inherent problem of existing battery technologies. Even with ICE energy conversion
efficiency figures of below 20%, the energy density (Joules/kg) of petroleum far
surpasses the energy density of any known battery technology [2]. Batteries are the
weak link in EVs at the moment [6]. The lack of a single reasonably priced energy
storage device that can simultaneously provide high power density and high energy
2
density for EVs has been the main stumbling block to the acceptance of EVs as the
main form of private and public transportation.
Presently the only viable solution to this problem is to combine a high energy
storage device such as an electrochemical battery or fuel cell with a high power
device such as an Electric Double Layer Capacitor (EDLC) or ultracapacitor or more
often called a supercapacitor [8]. In this work, we investigate the effect of
integrating the supercapacitor with the main power source (deep cycle lead acid
batteries) of an EV conversion. More so, at the University of Nottingham Malaysia
Campus, we have a supercapacitor pilot production plant which is dedicated to the
research and development of home grown supercapacitors and extensive materials
research. Concerted efforts have been made by the research group to investigate
the viable areas of application of supercapacitors such as consumer electronics,
mobile phones, solar and of course, the electric vehicle. The latter forms the basis
and also the genesis of the electric vehicle adventure at the University of
Nottingham Malaysia.
3
1.2 RESEARCH ISSUES ADDRESSED BY THIS WORK
The aim of this project is to address key issues that prevent electric vehicles from
being widely accepted and adopted by public and private transportation system.
Some of the research issues which may be regarded as problem statements are as
follows:
The low power density of EV batteries. In order for the EV to approach, or
try to approach, the performance of its conventional counterpart (ICE), the
battery or energy storage system should have similar or equivalent power
density as the ICE. The limitation of energy storage devices (batteries) to
deliver energy at high power rates during acceleration and also accept
energy during regenerative braking schemes. This is not to say that high
power density batteries do not exist, but at what cost and complexity for
practical use in a consumer electric vehicle.
Due to the haphazard nature of the load profile of an EV (traffic condition,
driver behaviour, etc), the battery suffers from inherent random high
discharge and charge (regenerative braking) rates as well. This is detrimental
to the shelf life of the battery
This led to the hybridization of batteries with supercapacitors as it becomes
impractical and cost ineffective to size a single energy storage system
(battery) for peak power demands several times higher than the mean
power demand. However, effective control strategies are still being
developed and redeveloped for energy management in a battery-
supercapacitor vehicle using various methods ranging from heuristics to
4
artificial neural networks to various learning algorithms. This power and
energy management poses as the most challenging task as it is to be done
without compromising the vehicles target performance.
The forecasting/prediction of driving load condition in advance is key to
efficient energy management systems for electric vehicles. An energy
management system for electric vehicles with multiple energy sources based
on the prediction of driving conditions or by analysing drive cycles.
5
1.3 RESEARCH OBJECTIVES
Based on the research issues highlighted above, the objectives of this research work
include the following:
To design and implement a small electric vehicle which will be solely
powered by a combination of deep cycle lead acid batteries and a
supercapacitor module. This will serve as a foundation for future research
work on electric vehicles at the university.
To design and implement an onboard data acquisition system to monitor
and log critical data from the electric vehicle.
To provide first hand insight into the interaction between lead acid battery
pack and supercapacitor module connected in parallel. This is achieved by
analysing novel real world drive data obtained by driving the vehicle in the
┌ミキ┗Wヴゲキデ┞げゲ ヮヴWマキゲWゲく
To investigate and justify the reported increase in battery life, range per
charge and vehicle acceleration by augmenting the battery pack with a
supercapacitor module. This is also achieved using real world drive data and
results.
6
1.4 RESEARCH SCOPE
When the initial research objectives for this project were conceived, the main aim
was to identify and propose an intelligent energy management system for electric
vehicles powered by a battery and supercapacitor combination. Emphasis was on
the optimum management of power and energy flow of multiple energy sources in
ラヴSWヴ デラ マWWデ デエW SWマ;ミS ラa デエW ┗WエキIノWげゲ ヮヴラヮ┌ノゲキラミ ;ゲ ┘Wノノ ;ゲ キデゲ ヴWケ┌キヴWマWミデ aラヴ
accessory loads. However, being the pioneer of the EV research team within the
research division, we had to create a solid platform or basis on which testing,
research and development can be carried out.
Due to inherent cost constraints, an ideal Hardware-In-the-Loop, (HIL) based EV test
bench could not be implemented. An experimental test vehicle would be
constructed; basically a conversion from a conventional ICE to electric. It would be
fitted with state of the art data acquisition system which monitors the interaction
between the energy and power sources as it tries to fulfil the load requirements of
the vehicle. In this case, the load on the propulsion system comes from 100% real
world driving. The only downside envisaged from this setup is the inability to
replicate the exact same driving pattern over a fixed route.
Experimental data obtained from this work is limited to the fixed driving route
within the university campus. Also the driving style of the driver could not be
controlled. However, the same driver was used in all experiments. Starting form a
pure battery driven vehicle, the energy system was then augmented with the
addition of a supercapacitor H;ミニく Aヮ;ヴデ aヴラマ マ;ミ┌a;Iデ┌ヴWヴげゲ キミaラヴマ;デキラミが デエWヴW キゲ
very limited experimental data on supercapacitor field testing. As such, the vehicle
7
provided a means to obtain unbiased empirical data to substantiate research claims
and also to serve as a test platform for further work.
1.5 THESIS CONTRIBUTION
This dissertation contributes to the emerging field of pure electric vehicles that are
powered by hybrid energy sources. This is achieved by providing first hand insight
into the interaction between lead acid battery pack and supercapacitor module
connected in parallel which is used to power a compact city car. Real world driving
data along with instantaneous voltages and currents from the energy sources
provide a novel and unique approach to obtaining unbiased empirical data to
substantiate research claims and also to serve as a test platform for further work.
The problems or challenges with having hybrid energy sources on-board a purely
electric vehicle rests heavily on the successful management of these sources in
order to meet the propulsion demand of the vehicle. As opposed to the general
consensus of designing and implementing complex control systems, this work takes
the basic parallel connection between battery and supercapacitor and investigates
the reported increase in battery life, range per charge and vehicle acceleration. This
is done with the hope that by analysing the natural power split between these two
devices in parallel, a practical, cost effective energy management system can be
developed in the future. Optimistically, several novel ideas will emerge from the
collaborative efforts of many small but progressive research contributions such as
this one.
8
1.6 THESIS ORGANIZATION
Chapter 2: Background and Literature Review に This chapter provides a brief
background on electric vehicles in terms of basic configurations, advantages and
disadvantages and also the hybrid vehicle configurations as well as related battery
chemistries. The basic structure and characteristics of the supercapacitor is tabled
out as well as its inherent pros and cons. A thorough literature review is presented
on multiple/hybrid energy sources for powering electric/hybrid vehicles with
particular emphasis on supercapacitors.
Chapter 3: Methodology に This chapter is divided into two parts. The first part
describes the equations governing the movement of an electric vehicle on a level or
inclined road surface. Using these equations, various software models were
developed in MATLAB Simulink © environment in order to simulate the
performance of the proposed experimental test electric vehicle conversion. The
second part describes the step by step design and construction of the experimental
test vehicle. Also, the traction and control wiring, sensor setup, data acquisition
system, on-board real time display of circuit parameters were laid out in detail.
Chapter 4: Results and Discussion に The results of the hardware implementation of
the battery supercapacitor combination are described and analysed. First hand
insight into the interaction between lead acid battery pack and supercapacitor
module connected in parallel was provided. Investigation and justification of the
reported increase in battery life, range per charge and vehicle acceleration by
augmenting the battery pack with a supercapacitor module was done.
9
Chapter 5: Conclusion and future work に Conclusion is presented here where major
accomplishments and research objectives achieved are summarised. The
possibilities for future work and improvements to be carried out will also be
discussed.
10
CHAPTER 2: BACKGROUND AND LITERATURE REVIEW
This chapter begins by providing a brief background on electric vehicles in terms of
basic configurations, advantages and disadvantages and also the hybrid vehicle
configurations. It proceeds to describe various battery chemistries that are used in
デラS;┞げゲ WノWIデヴキI ┗WエキIノW ;ミS ;ノゲラ デエW technical jargon associated with battery
technologies. An important term called state of charge (SOC) of a battery is
described as well as some methods to calculate it.
A エ┌ェW ヮラヴデキラミ ラa デエキゲ ヴWゲW;ヴIエ キゲ SWSキI;デWS デラ┘;ヴSゲ キミ┗Wゲデキェ;デキミェ ; IWヴデ;キミ SW┗キIWげゲ
use or relevance in the ever growing electric/hybrid vehicle industry; the
supercapacitor or ultracapacitor or electric-double-layer capacitor (EDLC). The basic
structure and characteristics of the supercapacitor is tabled out as well as its
inherent pros and cons. One feature which sticks out is its high power density which
is very complimentary to present day battery technologies.
A thorough literature review is presented on multiple/hybrid energy sources for
powering electric/hybrid vehicles with particular emphasis on supercapacitors. The
challenges faced by various researchers in this field are outlined which are
associated with energy management of the multiple energy and power sources.
Various combination algorithms have been proposed and results shown ranging
from fuzzy logic, neural networks, genetic algorithms to support vector machines.
Finally, the support vector machine (SVM) which has been identified as one of the
leading classifiers in machine learning is described.
11
2.1 BACKGROUND
2.1.1 Electric Vehicles
Electric vehicles (EV) are propelled by an electric motor (or motors) powered by a
ゲデ;Iニ ラa ヴWIエ;ヴェW;HノW H;デデWヴキWゲく Iミ EVげゲが デエW キミデWヴミ;ノ IラマH┌ゲデキラミ WミェキミW ふICEぶ キゲ
replaced by an electric motor which gets its power from a controller circuit.
The first EVs of the 18th
century used non-rechargeable batteries and by the end of
the 19th
IWミデ┌ヴ┞が マ;ゲゲ ヮヴラS┌Iデキラミ ラa ヴWIエ;ヴェW;HノW H;デデWヴキWゲ マ;SW EVげゲ a;キヴノ┞
common[1]. At the start of the 20th
IWミデ┌ヴ┞が EVげゲ ゲWWマWS デラ HW ; ゲデヴラミェ IラミデWミSWヴ
for future road transport as ICE vehicles were at the time unreliable, extremely
pollutive, and needed to be cranked manually to start.
HラヮW aラヴ EV ┘;ゲ S;ゲエWS ┘エWミ キミ デエW W;ヴノ┞ ヱΓヰヰげゲが IエW;ヮ ラキノ ┘;ゲ ┘キSWノ┞ ;┗;キノ;HノW
;ミS ; ゲWノa ゲデ;ヴデWヴ aラヴ ICEげゲ ┘;ゲ SW┗WノラヮWSく IC Wngines proved a more attractive
option for powering vehicles and rechargeable batteries were ironically adopted for
ゲデ;ヴデキミェ ICEげゲ ぷ2].
Despite the above problems, electric vehicles always found its use in delivery vans,
warehouse vehicles and golf carts etc. [2]. Environmental issues may well be the
deciding factor in the adoption of EVs for town and highway uses coupled with the
fact that tremendous technological developments have taken place in vehicle
design, improvement to rechargeable batteries, motors and controllers. It is evident
that electric vehicles have come to stay this time around.
12
FIG 1: The Basic Electric Vehicle Configuration (Source: HowStuffWorks[79])
Figure 1 represents the basic principle of operation of an electric vehicle. The
accelerator pedal is connected to a potentiometer which determines how much
power the controller must deliver to the motor. The electric motor can be either ac
or dc type.
A dc motor and controller system tends to be simpler and less expensive to install.
They can also be overdriven i.e. accept up to ten times its rated power for a few
seconds. This is useful for providing short bursts of acceleration. However the
motor is bound to heat up [3].
On the other hand ac systems allow for the use of any industrial 240V three phase
motor which is readily available. Also, regenerative braking is possible as ac motors
can be turned into generators to recover braking energy. It is without doubt that ac
controllers are more complicated due to the fact that dc has to be converted to
three phase ac using inverters.
13
Advantages over ICEs
I. They are more energy efficient; more than 50% of the chemical energy in a
battery is converted to electrical energy which is used to power the wheels
whereas only around 20% of the energy in petrol/diesel is converted to
electrical energy
II. Since no combustion takes place, there are no tail pipe emissions
III. Electricity is a domestic source of energy; thus EVs can be recharged easily
and conveniently.
IV. Quiet smooth operation and requires less maintenance as compared to ICE
counterpart
The downside of EVs is usually battery related. Batteries are the weak link in EVs at
the moment.
I. The driving range is much less ; the battery needs to be recharged more
often
II. Usually recharge time for the batteries is too long (4-8 hours).
III. Battery packs are expensive and need to be replaced more number of times
IV. Battery packs are heavy, heavy and take up a lot of space.
Researchers all over the world are working on improved battery technologies to
make up for the disadvantages mentioned above. The results of this research will go
a long way in determining whether EVs are fully adopted in the future.
Electric cars are mechanically much simpler than both gasoline cars and fuel-cell
cars. There is no motor oil, no filters, no spark plugs, and no oxygen sensors. The
motor has one moving part, there is no clutch, and the transmission is much
14
simpler. Due to regenerative braking, even the friction brakes will encounter little
wear. The only service that a well-designed electric car will need for the first
100,000 miles is tire service and inspection [3].
Until an electric car manufacturer achieves high enough sales to approach a
ェ;ゲラノキミW I;ヴ マ;ミ┌a;Iデ┌ヴWヴげゲ ┗ラノ┌マW WaaキIキWミIキWゲが WノWIデヴキI I;ヴゲ ┘キノノ ミWWS デラ IラマヮWデW
on other grounds besides price. Aside from the obvious emissions advantage, there
is another way that an electric car can vastly outperform a gasoline car に in a word,
デラヴケ┌Wく A ェ;ゲラノキミW WミェキミW エ;ゲ ┗Wヴ┞ ノキデデノW デラヴケ┌W ;デ ノラ┘ ヴヮマげゲ ;ミS ラミノ┞ SWノキ┗Wヴゲ
reasonable horsepower in a narrow rpm range. On the other hand, an electric
motor has high torque at zero rpm, and delivers almost constant torque up to about
6,000 rpm, and continues to deliver high power beyond 13,500 rpm. This means
that a performance electric car can be very quick without any transmission or
clutch, and the performance of the car is available to a driver without special driving
skills [3].
Fig 2: Emissions: EV vs. ICE (Source: Sustainability-ed.org.uk [80])
15
The graph in the figure 2 compares the emissions from petrol cars and electric cars
using electricity generated in conventional power stations. Overall, there are fewer
emissions from electric cars. This could help to reduce global warming and also
improve the air quality in our cities. According to the U.S. Environmental Protection
Agency (EPA), vehicle emissions currently contribute between one third and one
half of the total U.S. atmospheric burden of three major pollutants: carbon
monoxide (CO), nitrogen oxides (NOx), and hydrocarbons (HC) [4]. Their impact is
even greater in many U.S. urban areas. This comparison does not include emissions
from ICEs due to lack of maintenance. One school of thought argues that even
though EVs do not produce tailpipe emissions, the electricity which is used to
charge them comes from a power plant that burns fuel. Many researchers have
studied this problem, and the general conclusion is that, including the power plant
emissions, the electric vehicle is much less polluting. How much cleaner it is,
depends on what energy source the power plant uses, coal being the dirtiest. But if
you can transform the grid to run from a renewable source, then we may have
something exciting to talk about. However, we're not there yet.
Although this report is focused on the pure electric vehicle or sometimes called
battery electric vehicle (BEV), it was mandatory to sデ┌S┞ ;Hラ┌デ HEVげゲ ┘エキIエ ; デヴ;SW-
ラaa HWデ┘WWミ EVげゲ ;ミS ICE WミェキミWゲく Aノゲラ ; ノラデ ラa ヴWゲW;ヴIエ ノキデWヴ;デ┌ヴW エ;ゲ HWWミ
ヮ┌HノキゲエWS ラミ WミWヴェ┞ マ;ミ;ェWマWミデ ゲデヴ;デWェキWゲ aラヴ HEVげゲ ┘エキIエ Iラ┌ノS HW ;ヮヮノキI;HノW デラ
multiple energy sources on board an electric vehicle.
16
2.1.2 HEV drive configuration
HEVげゲ I;ヮ;Hキノキデ┞ ラa ヴWS┌Iキミェ a┌Wノ Iラミゲ┌マヮデキラミ ;ミS Wマキゲゲキラミゲ ┘エキノW マ;キミデ;キミキミェ デエW
same high performance as conventional vehicles lies in a key factor. This factor is
being able to operate the ICE under near optimal conditions [5].
Fig. 3: Series Hybrid Drive (Adapted from US DoE Office of Transport [5])
Series Hybrid Drive
The internal combustion engine drives an alternator to generate electricity that
drives the traction motor (wheels) directly or it can be stored in a battery. The
traction motor can be supplied solely from the battery (zero emissions mode), from
the engine or from a combination of both. In the event that the battery discharges
below a certain level, the engine turns on and recharges the battery. It should be
noted that the internal combustion engine is not mechanically connected to the
wheels thus less dependent on the vehicles changing power demands. In this drive
configuration, the ICE can operate at optimum conditions but the efficiency of the
electrical system suffers due to series connected components.
17
Fig.4: Parallel Hybrid Drive (Adapted from US DoE office of Transport [5])
Parallel Hybrid Drive
In parallel hybrid drive configuration, both the internal combustion engine and the
electric motor are mechanically connected to the transmission system i.e. the
wheels. A parallel HEV typically has two power paths; either the ICE or the electric
motor or both can be used to provide the power to turn the wheels. A computer
controls the electric motor depending on the power needs of the vehicle. For
example, the electric system may be used for short trips while for longer trips the
ICE ┘ラ┌ノS ヮヴラ┗キSW デエW ┗WエキIノWげゲ マ;キミ ヮラ┘Wヴ ┘キデエ デエW WノWIデヴキI ゲ┞ゲデWマ ゲ┌ヮヮラヴデキミェ キミ
periods of sudden acceleration or other peak power demands. A drawback however
is that the ICE load is more dependent on the actual driving situation.
18
Combined Switched Hybrid drive
The combined switched drive allows for a purely series or parallel drive by the
activation of the clutch. When driving at low speeds i.e. within the city, the series
regime can be used and the ICE operated at optimum condition. For inter-city
driving which requires higher output from the ICE, the parallel regime can be
employed.
Power Splitting Systems
Propulsions with electric power splitting are based on ICE power splitting into two
parts. One part is converted in electric power in generator, which supplies traction
motor mechanically connected to drive wheels and remaining part is transmitted by
electromagnetic forces in the air gap to the wheels mechanically without losses in
electric machines. The splitting device can be realized mechanically by planetary
gear or electrically by a special electric generator.
Although HEVs are by now an industrial standard in the automotive industry
experiencing an exponential growth, they will be no further discussion on them in
this report. The sole focus will instead be on pure electric vehicles (no IC engine at
all), which consists of the use of more than one type of electrical energy storage
device. Fuel cell vehicles (FCV) and the related problems of hydrogen production
and storage, although recognized as being an extremely interesting technology, are
also beyond the scope of this thesis. However, it is pointed out that most of the
concepts related to hybridization discussed in the following will readily apply to fuel
cell vehicles, since in virtually every implementation the slow dynamics of the fuel
19
cell require the use of some kind of additional power source, like a battery or a
supercapacitor bank.
2.1.3 Battery Technologies for the Electric Vehicle
Technological advances in the automotive industry have led to the demand for
マラヴW ヮラ┘Wヴ キミ デラS;┞げゲ automobile. This has heaped a lot of pressure on battery
manufacturers to produce batteries which can meet the demands of power hungry
applications (anti-lock braking system, air conditioning, power steering, etc.) as well
as improve fuel efficiency.
The classic electric vehicle has only one energy storage unit; the battery. It is usually
that component with the highest cost, weight and volume. Likewise, in hybrid
vehicles the battery must continually supply or store electrical energy. The battery
is of utmost importance; the key and enabling technology to the electric vehicle
revolution. The basic requirements for a battery used in electric vehicles are as
follows:
(i) Low cost: the ultimate aim of the automotive industry
(ii) Lowest possible weight to reduce load on drive train
(iii) High cycle-life: ideally to last the lifetime of the vehicle, practically for a
reasonably long period of time.
(iv) High-energy density can be attained with one charge to provide a long
range or mileage
(v) High power delivery and fast recharge capability: this is a very desirable
requirement as batteries could be fully recharged in 5-10mins; the time
20
it takes to refill a gas tank. Also, energy saving schemes such as
regenerative braking can be easily incorporated. Energy can be captured,
stored and delivered as soon as required.
(vi) Safety is also a big concern in terms of overcharging, over heating etc.
(vii) Wide acceptance as a recyclable battery from the environmental
standpoint
At the present time, the battery types being considered and developed for electric
vehicles are relatively few, namely lead-acid, nickel metal hydride, lithium-ion and
lithium polymer, and sodium nickel metal chloride (ZEBRA). Each of these battery
types has some advantages and disadvantages and unfortunately none of them are
attractive in all respects for electric vehicles. The main parameters that specify the
behaviour and performance of a battery are given in the following section:
Some Common Battery Parameters
Ampere-hour Capacity: This parameter, measured in coulombs (C) but usually
expressed in Ampere-hour (Ah) capacity, is the total charge that can be discharged
from a fully charged battery under specified conditions. The rated Ah capacity is for
a new battery under the conditions predefined by the manufacturer. Watt-hour
(Wh) or kWh is also used to represent battery capacity.
噺 抜
For example, a 48V 225 Ah pack will have a rate capacity of 10.8 Kwh.
C-rate: The charge and discharge of a battery is measured in C-rate. The nominal C-
rate is used to represent a charge or discharge rate equal to the capacity of a
21
battery in one hour i.e. 1C. Most portable batteries, with the exception of the lead
acid, are rated at 1C. A 6V, 225Ah Trojan T105 lead acid battery is rated for 20hour
discharge i.e. C20 or C/20 or 0.05C. This means that if the battery were to be
discharged over a period of 20hours, then a current draw of 11.25A is
approximately expected.
Specific Energy (W/kg): This is the nominal battery energy per unit mass; a
characteristic of the battery chemistry and packaging. Along with the energy
consumption of the vehicle, it determines the battery weight required to achieve a
given electric range.
Specific Power (W/kg): This is the maximum power per unit mass which is available.
It determines the battery weight required to achieve a given performance target.
Internal Resistance: Internal resistance is the overall equivalent resistance within
the battery. It is different for charging and discharging and may vary as the
operating condition changes. As internal resistance increases, the battery efficiency
decreases and thermal stability is reduced as more of the charging energy is
converted into heat.
State of Charge (SOC): It is a dimensionless parameter defined as the remaining
capacity of a battery and it is affected by its operating conditions such as load
current and temperature. For battery electric vehicles, it is synonymous to the fuel
gauge in a conventional vehicle [6]. There are several methods of estimating the
SOC of a battery such as voltage method, specific gravity, coulomb counting or
current integration, impedance spectroscopy and newer methods like quantum
magnetism. The SOC in simple terms is defined as:
22
噺 迎結兼欠件券件券訣 系欠喧欠潔件建検 迎欠建結穴 系欠喧欠潔件建検
SOC is a critical condition parameter for battery or energy management. Accurate
gauging of SOC is very challenging, but the key to the healthy and safe operation of
batteries. In the sections to come, various techniques from literature have been
proposed on estimating this parameter. SOC is generally calculated using current
integration to determine the change in battery capacity over time as described in
the equation below:
ッ 噺 岫 岻 伐 岫建待岻 噺 な 畦月 系欠喧欠潔件建検 豹 件岫酵岻穴酵痛痛待
Depth of Discharge (DOD): DOD is used to indicate the percentage of the total
battery capacity that has been discharged. For deep-cycle batteries, they can be
discharged to 80% or higher of DOD. Typically, it is desired that the DOD of a battery
is kept within appropriate safe limits as specified by the manufacturers.
噺 な 伐
State of Health (SOH): Tエキゲ ェキ┗Wゲ ;ミ キミSキI;デキラミ ラa デエW H;デデWヴ┞げゲ ヮヴWゲWミデ IラミSキデキラミ キミ
terms of its present energy capacity as compared to the rated energy capacity when
it was new [7].SOH is an important parameter for indicating the degree of
performance degradation of a battery and for estimating the battery remaining
lifetime.
噺 畦訣結穴 継券結堅訣検 系欠喧欠潔件建検迎欠建結穴 継券結堅訣検 系欠喧欠潔件建検
23
Cycle Life (number of cycles): Cycle life is the number of dischargeにcharge cycles the
battery can handle at a specific DOD (normally 80%) before it fails to meet specific
performance criteria.
Battery Management System (BMS).: It is a network of sensors, controller,
communication, and computation hardware with software algorithms designed to
decide the maximum charge/discharge current and duration from the estimation of
SOC and SOH of the battery pack.
A summary of the main battery technologies used in electric vehicles available
today are as follows [8]:
Lead-acid
I. Lead as negative electrode, lead oxide as positive electrode and dilute
sulphuric acid as electrolyte or in more advanced forms as gels
II. Ruggedness, low cost, inherent safety, and temperature tolerance
III. It is a mature technology
IV. It reached the mass production stage a very long time ago
V. The valve regulated lead-acid (VRLA) battery is maintenance-free and has
good cyclability even at deep discharge; two types absorbed glass mat
(AGM) based and gel based batteries
VI. Limited depth of discharge DOD
VII. Limited life cycle when deep discharged
VIII. Low specific energy and power due to heavy lead collectors
IX. Not environmentally friendly to dispose
24
Ni-MH (Nickel-Metal Hydride)
I. Nickel Hydroxide as positive electrode, alloy as a negative electrode and an
alkaline solution as electrolyte
II. Most currently available hybrid electric vehicles use this chemistry as energy
storage system
III. High volumetric energy and power as this technology has experienced great
advances in the past 15yrs
IV. Safe operation at high voltages
V. Environmentally friendly and recyclable
VI. Longer cycle life when compared to lead acid as it has a reasonable
tolerance over abusive overcharge and discharge
VII. Wider Operating temperature due to better thermal properties
VIII. The cost is a huge problem as the high price of the raw material, mainly
nickel, will not come down in mass production
Nickel Zinc (NiZn)
I. Anode is Zinc, cathode is from Nickel and electrolyte is an alkaline solution
II. Specific energy and power greater than lead acid chemistry
III. NiZn batteries do not use mercury, lead, or cadmium, or metal hydrides that
are difficult to recycle. Both nickel and zinc are commonly occurring
elements in nature, and can be fully recycled.
IV. suffer from poor lifetime due to the fast growth of dendrites
Nickel Cadmium (NiCd)
I. Nickel oxide as positive electrode (cathode), cadmium compound as
negative electrode (anode) and potassium hydroxide as electrolyte
25
II. Where energy density is important, NiにCd batteries are now at a
disadvantage compared with nickelにmetal hydride and lithium-ion batteries
III. Smaller and lighter than lead acid equivalents
IV. Can endure high discharge rates without any significant loss of capacity or
damage
V. High cost of cadmium and nickel
VI. Cadmium is an environmental hazard, and it is highly toxic to all higher
forms of life hence requires special care during disposal
Lithium Ion Batteries; Next Generation of Electric Vehicles
Lithium is the lightest of all metals, has the greatest electrochemical potential and
provides the largest energy density for weight. In 1991, the Sony Corporation
commercialized the first lithium-ion battery. Lithium-ion batteries are incredibly
popular these days. You can find them in laptops, PDAs, cell phones and iPods [9].
Lithium-ion is a low maintenance battery, an advantage that most other chemistries
cannot claim. There is no memory effect and no scheduled cycling is required to
prolong the battery's life. In addition, the self-discharge is less than half compared
to nickel-cadmium, making lithium-ion well suited for modern fuel gauge
applications. Lithium-ion cells cause little harm when disposed [10].
Despite its overall advantages, lithium-ion has its drawbacks. It is fragile and
requires a protection circuit to maintain safe operation. Built into each pack, the
protection circuit limits the peak voltage of each cell during charge and prevents the
cell voltage from dropping too low on discharge. This is because a total discharge of
the Li-ion battery can irreversible as it may not be able to be recharged anymore.
26
Thus what will happen if a vehicle with a Li-ion battery pack is unused for say one
month?
In addition, the cell temperature is monitored to prevent temperature extremes. Li-
on does not fare well in very cold climates. Below zero temperature seriously
impairs its battery capacity, or ability to provide its full discharge capacity at lower
temperatures. Aging is a concern with most lithium-ion batteries and many
manufacturers remain silent about this issue. Some capacity deterioration is
noticeable after one year, whether the battery is in use or not. It should be noted
that other chemistries also have age-related degenerative effects. Manufacturers
are constantly improving lithium-ion. New and enhanced chemical combinations are
introduced every six months or so. With such rapid progress, it is difficult to assess
how well the revised battery will age.
Lithium Iron Phosphate (LiFePO4) is proving to be a breakthrough for EV batteries.
Based upon lithium ion technology, LiFePO4 batteries offer many advantages over
lithium cobalt dioxide (LiCoO2) batteries which are commonly used in laptops, mp3
players and cell phones. In EVs, LiFePO4 batteries offer greater range, power and
safety. They provide full power until they are completely discharged, and recharge
in just 2.5 hours. LiFePO4 chemistry is also environmentally friendly ね キデげゲ デエW ノW;ゲデ
toxic of all the battery types. Lithium Nickel Manganese Cobalt Oxide technology is
the preferred candidate for EVs [11], as it provides an excellent trade-off between
specific energy and power, cost, performance, life span and cost.
The figure 5 below shows the energy densities in Wh/kg of various battery
technologies.
27
Fig. 5: A Comparison of Energy densities of various batteries (Source: Battery
University [81])
2.1.4 Supercapacitors/ Ultracapacitors/ Electric Double Layered
Capacitors (EDLC)
As the name implies, a supercapacitor/ultracapacitor/EDLC (names used
interchangeably) is a capacitor with capacitance greater than any other, usually in
excess of up to 4000 Farad. They possess a greater specific energy density (Wh/kg)
than conventional electrolytic capacitors and a higher specific power density (W/kg)
than most battery technologies. Supercapacitors do not have a traditional dielectric
material like ceramic, polymer films or aluminum oxide to separate the electrodes
instead a physical barrier made of activated carbon [12]. A double electric field
which is generated when charged, acts as a dielectric. The surface area of the
activated carbon is large thus allowing for the absorption of large amount of ions
[13]. Figure 3 below is the basic structure of a supercapacitor.
28
Fig 6: Electric Double Layer Capacitor Structure (Source: EAMEX Capacitor [82])
Activated carbon is impregnated with electrolyte. Positive and negative charges are
formed between the electrodes and the impregnate. The magnitude of voltage
where charges begin to flow is where the electrolyte begins to break down. This is
called the decomposition voltage. Supercapacitors have a low energy density of less
than 15Wh/kg but a very high power density of up to 4000W/kg and capacitance
values of thousands of Farads are possible [14].
As with conventional capacitors, the capacitance (C) of a supercapacitor is simply
the ratio of the stored charge (Q) to the applied voltage (V). Also C is directly
proportional to the surface area A of each electrode and inversely proportional to
the distance D between the electrodes. This is described in the equation below:
系 噺 香墜綱追 畦経
29
香墜 is the dielectric constant of free space and 香追 is the dielectric constant of the
insulating material between the electrodes or the electrically non permeable
separator. They are constants which depend on the type of material being used.
Since the surface area of the activated carbon electrodes is very high; 1000-2300
m2/g, as compared to the distance between them; 10 Angstroms or less, the
capacitance can be potentially very high [15]. Their voltage ratings are typically low
(2.5V to avoid electrolysis of the electrolyte) thus they are combined in
modules/banks (series connection) to handle higher voltages.
The energy, E stored in a supercapacitor is directly proportional to its capacitance:
継 噺 なに系撃態
The greatest factor in determining the supercapacitor performance is the energy
lost through internal resistance, represented by an ESR (equivalent series
resistance). Thus reducing ESR increases the power density [16].
Variations in individual cell resistance and capacitance result in uneven voltage
distribution in a bank causing destruction of the cell. A circuitry which is able to
spread the available charge evenly between the capacitors is necessary for a
supercapacitor bank. This is called a voltage balancing or equalization circuit. There
are two reasons for an imbalance of voltages in a serial string of supercapacitors: (i)
deviations from the nominal capacitance of the capacitors and (ii) deviations in self
discharge performance i.e. deviations in the leakage current.
30
Advantages of Supercapacitors
I. Cell voltage determined by the circuit application not limited by cell
chemistry
II. Very high cell voltages possible
III. High power density
IV. Can withstand extreme temperatures
V. Simple charging methods
VI. Very fast charge and discharge
VII. Long life cycle
VIII. Low impedance
Disadvantages/Shortcomings
I. Linear discharge voltage characteristic prevents use in some applications
II. Power only available for very short duration (short bursts of power)
III. Low capacity
IV. Low energy density
V. Voltage balancing required when banking
VI. High self discharge rate
The EDLC stores charges in a different way as compared to the conventional
electrolytic capacitor [17]. Thus traditional models are inadequate to describe the
EDLC although the basic equations for capacitors still apply. The double layers
formed on the activated carbon surfaces of a supercapacitor can be modeled as a
series of parallel RC circuits as shown in the figure below.
31
Fig 7: Electric Double Layer Modeled as Series of Parallel RC Circuits
WエWヴW Rヱが Rヲが Rン ぐ Rミ ヴWヮヴWゲWミデ デエW ゲWヴキWゲ ヴWゲキゲデ;ミIW ;ミS Cヱが Cヲくくくが Cミ ;ヴW デエW
electrostatic capacitances of the activated carbons. When voltage is applied,
current flows through each of the RC circuits.
Figure 8 below represents the first order model of the supercapacitor which
comprises of four ideal circuit elements: a capacitance C, a series resistor Rs, a
parallel resistor Rp, and a series inductor L. Rs is called the equivalent series
resistance (ESR) and contributes to energy loss during capacitor charging and
discharging i.e. current flow. Rp simulates energy loss due to capacitor self-
discharge, and is often referred to as the leakage current resistance. Inductor L
results primarily from the physical construction of the capacitor and is usually small.
Fig 8: 1st
Order Model of the Supercapacitor
In actuality, Supercapacitors exhibit a non-ideal behaviour due to the porous
material used to form the electrodes that cause the resistance and capacitance to
32
be distributed such that the electrical response mimics transmission line behaviour.
This is shown in the figure below.
Fig 9: Non Ideal (transmission line) Model of the Supercapacitor
The most common supercapacitor applications are memory backup and standby
power. In some special applications, the supercapacitor can be used as a direct
replacement of the electrochemical battery. Additional uses are filtering and
smoothing of pulsed load currents [18]. A supercapacitor can, for example, improve
the current handling of a battery. During low load current, the battery charges the
supercapacitor. The stored energy then kicks in when a high load current is
requested [19]. This enhances the battery's performance, prolongs the runtime and
even extends the longevity of the battery. The shortcomings of a supercapacitor
may render them unsuitable as the primary source of power in an EV or HEV.
However their unique features make them ideal for temporary energy storage such
as storing energy via regenerative braking and also for providing a booster charge in
response to sudden power demands in vehicles [20]. Thus the idea of combining the
high power density features of a supercapacitor with the existing high energy
density of a battery is very promising and thus a subject of intense research
nowadays.
33
The best way to compare the capabilities of the storage devices mentioned above is
by a Ragone chart. The values of energy density (in Wh/kg) are plotted versus
power density (in W/kg). Both axes are logarithmic, which allows comparing
performance of very different devices (for example extremely high, and extremely
low power). From the chart below, it will be observed that the batteries are in the
region of high energy density, low power density and longer discharge time while
the ultracapacitor is in the region of high power density, low energy density and
very fast discharge time. It should be noted that the diagonal lines across the chart
represents discharge time.
Fig 10: A Ragone Chart (Source: Don Tuite [83])
34
Chemistry Nom. Voltage
(V)
Energy
Density
(Wh/kg)
Power Density
(W/kg)
Cycle life
Lead-acid 2 30-40 180 чϴϬϬ
Ni-Cd 1.2 ~50 150 чϱϬϬ
Ni-Mh 1.2 55-80 400-1200 чϭϬϬϬ
Li-Phosphate 3.2~3.3 80-125 1300-3500 чϮϱϬϬ
Li-ion 3.6 80-170 800-2000 чϭϱϬϬ
Li-Manganese 3.7 110-130 чϮϬϬϬ
Li-polymer 3.7 130-200 1000-2800 чϭϱϬϬ
Supercapacitor 2.5 or 2.7 5-10 4000-10,000 шϭ͕ϬϬϬ͕ϬϬϬ
Table 1: Specific battery figures in comparison with supercapacitor figures (Sources:
[21]-[28]).
The table above clearly shows the disparity between batteries (all chemistries) and
the supercapacitor. Batteries have an energy density varying from 30-200 Wh/kg
while the supercapacitor has only 5% of that energy density. On the other hand, the
power density of the supercapacitor is 4 to 10 times greater than that of any known
battery chemistry to date. Also it can boast of 500 times more calendar life than its
battery counterparts. Researchers have typically classified batteries and
supercapacitors as electrochemical devices. However operating principles of both
these devices are different which make their characteristics highly different. It
becomes very obvious that a synergy between these two devices (battery and
supercapacitor) will bring about an overall increase in efficiency and power delivery.
It should be noted that this hybrid combination is not limited to electric or hybrid
vehicles as it can be substituted for any conventional energy storage system which
has a fluctuating load profile. However the aim of this research work is to be
centred around battery electric vehicles only.
35
2.2 STATE-OF-THE-ART ON HYBRID/MULTIPLE ENERGY
STORAGE SYSTEMS FOR ELECTRIC VEHICLES
Usually when two or more energy sources are involved in a hybrid energy storage
system for an electric vehicle, these sources can be distinguished by their energy
storage and power delivery capacities respectively. For a pure electric vehicle,
sources with high energy density would be considered as the main energy source
such as battery packs and fuel cells. As such, they can solely power a vehicle from
point A to point B even though the efficiency figures could be low. To boost these
figures to a reasonable level, an auxiliary energy source synonymous with high
power density or delivery is usually utilized. Popular choices for this source include
high power batteries and supercapacitors. It should be noted that a vehicle may not
be able to run solely on this auxiliary source albeit for a few kilometres only. Due to
デラS;┞げゲ ヴWケ┌キヴWマWミデ ラa maximum regenerative braking energy recapture, power
batteries being a chemical energy conversion process are not as dynamic enough
for this purpose. Supercapacitors on the other hand remain the only viable
alternative for a powerful yet dynamic auxiliary source which is able to compliment
the already stable main energy source. This provided a good motivation for this
research work which aims to integrate and analyse the effects of supercapacitors as
peak power buffer system for electric vehicles.
Engineers generally address peak power needs by designing the primary energy
source to the size needed to satisfy peak demands, even if those demands occur for
only a few seconds. Sizing an entire system for peak power needs, rather than for
the average power requirement may prove costly and inefficient. Such systems can
36
be significantly improved by storing electrical energy from a primary energy source
such as a battery in a secondary energy storage device, and then delivering that
energy in controlled high power bursts when required. Such high power delivery
provides electrical systems with dynamic power range to meet peak power
demands for periods of time ranging from fractions of a second to several minutes.
Batteries are not designed to provide bursts of power over many hundreds of
thousands of cycles. Supercapacitors perfectly meet this requirement. The general
scope of an energy management system for multiple sources is described in the
figure below.
Fig 11: General Overview of an Energy Management System (Source: EEtimes Design
[84])
The figure above represents what the EMS of an electric vehicle should effectively
do i.e. main energy source (battery) for average power demand, auxiliary energy
37
source (supercap) for peak power demand and energy capture (regenerative
braking) schemes.
The issue of power arbitration or sharing between multiple sources to satisfy the
load requirement may be seen in another light as in the figure below which was
W┝ヮノ;キミWS キミ LくC Rラゲ;ヴキラげゲ PエD デエWゲキゲ ぷヵヴへ
Fig 12: Power arbitration concept for multiple energy sources (Source: L.C Rosario
[54])
The figure above can be explained in very simple terms as follows:
- There are usually N storage devices
38
- W: weight factor for adjusting the rate at which energy is being drawn
(power) from each source.
- пぎ ;ノェラヴキデエマ デラ IララヴSキミ;デW デエW ヮラ┘Wヴ aノラ┘ H┞ S┞ミ;マキI;ノノ┞ ┗;ヴ┞キミェ デエW
weight factors in response to the ever dynamic driving load. Also taking into
consideration the system constraints such as depletion levels or limits of the
energy storage systems
- The load requests fluctuate through a drive cycle in the case of an electric
vehicle
During the course of this research project, a few hundred literatures was collected
including textbooks, journal papers, conference papers, application notes, patents
and standards. A summary of work conducted by research groups in the area of
energy/power management for electric vehicles with hybrid energy sources is
presented in the sections below. The works presented range from a simple parallel
connection, series configurations to complicated intelligent energy management
systems using some sort of machine learning theory. It should be noted that in all
literature surveyed, the supercapacitor was acting as the auxiliary source while a
battery pack or fuel cell, as the case may be, was the main energy source.
The most basic method of combining a supercapacitor with a battery pack is a
simple parallel configuration. The performance of a batteryにsupercapacitor hybrid
power source under pulsed load conditions was analytically described using
simplified models in this work [29], to reveal potential gains in peak power,
reduction of internal losses and extension of discharge life. A basic circuit shown in
39
the figure below, and its thevenin equivalent in frequency domain was used to
perform a theoretical analysis.
Fig 13: Supercapacitor-battery parallel circuit and its Thevenin equivalent in
frequency domain (Source: Dougal et al [29])
CC is the supercapacitor nominal capacitance and RC, its equivalent series resistance
or ESR. Vb キゲ デエW H;デデWヴ┞げゲ キSW;ノ ┗ラノデ;ェW ;ミS キミデWヴミ;ノ ヴWゲキゲデ;ミIW キゲ Rb. Based on the
circuits above, the load, battery and supercapacitor currents were simulated with
particular values of resistance and capacitance with pulse frequency of 0.1 Hz and
load duty ratio of 0.1. It was observed that the current supplied by the capacitor
during the load, on-state relieves peak stresses on the battery and therefore
influences the performance of the entire system is a positive way. A number of
factors affecting the amount of current supplied by the supercapacitor were
identified as bank size and configuration, pulse rate and duty ratio of the load. The
peak power enhancement capability of the supercapacitor was theoretically
analysed as the direct result of its current sharing capability; small series resistance
which can be obtained cascading more capacitors in parallel than in series. This
relates to the supercapacitor bank size and configuration. The result of this analysis
40
was used to optimize a system design case study and it resulted in a power loss
saving of 74% from the battery and also an increase in run time of the entire system
by 8.4 minutes.
An insightful piece of practical research case study on the integration of batteries
and supercapacitors for electric vehicle propulsion would be [30], carried out at
Massachusetts Institute of Technology, MIT. It was a student driven project aimed
at powering a go-kart with a combination of lead acid batteries and supercapacitor
module in a series configuration. Figure [13] below describes this configuration.
Fig 14: Battery + Supercapacitor Series Configuration (Source Colton.S [30])
In this configuration, the voltage present on the supercapacitor reduces the current
demand from the battery at a given load current. The battery pack is connected to
the traction motor via a half bridge converter which essentially acts as an isolation
switch; pulse width modulation to turn on and off the switching circuit. A high
power by-pass diode enables the go-kart to be powered solely by the battery pack
while the capacitor contactor switch enables a boost mode whereby the
supercapacitor is in series with the motor and contributes to the load demand. Also,
regenerative braking is solely channelled into the supercapacitor. Positive attributes
41
of this configuration as observed on the go-kart implementation included the
following:
Small supercapacitor and its entire voltage range utilized at a lower cost
when compared to the traditional parallel configuration
No need for external inductors as all the switched current passed through
the motor windings
Off-the-shelf dc to dc unidirectional converters may be used to connect the
battery to the traction motor.
The disadvantages highlighted in this work being no direct power transfer between
the two power sources i.e. the battery cannot charge the supercapacitor and vice
versa. Also the regenerative braking energy cannot be used to charge the battery
pack.
A direct parallel connection of the supercapacitor across the battery terminals does
reduce transient currents in and out of the battery according to Pay et al. Figure 15
below describes the direct parallel connection in which the supercapacitor has to be
prWIエ;ヴェWS デラ デエW H;デデWヴ┞げゲ デWヴマキミ;ノ ┗ラノデ;ェW aキヴゲデく Tエキゲ マ;ニWゲ キデ キマヮラゲゲキHノW デラ
control the power flow in and out of the supercapacitor bank because its terminal
voltage is tied to that of the battery at all times. In effect, current division between
these two devices solely rests on the value of their internal resistances [31]. An
Experimental setup consisting of valve-regulated lead acid (VRLA) batteries to make
up a total of 336VDC, 170Ah and a string of supercapacitors each 2.5V 2500F
totalling 375VDC 16.67F were used for validation. Results such as a 40% reduction
in battery pack current and 30% improvement in dc bus voltage regulation were
42
obtained. However, it was observed that the state of charge (SOC) of the
supercapacitor dropped considerably after an acceleration.
Fig 15: Direct Parallel Connection of Battery bank (b) with Supercapacitor Bank (c)
supplying a load. (Source Pay.S et al [31])
Authors Gao, Dougal and Liu, went a step further with their original work [29], to
produce a comparison between passive (direct paralleling) and active (via a
converter) configurations between a battery and supercapacitor (see figure 15
below) using a software simulation package called the Virtual Test Bench (VTB) [55].
A hardware implementation was also carried out using real time dSPACE controller,
SonyUS18650 lithium ion battery pack, Maxwell PC100 supercapacitor and a
programmable pulsed load.
43
Fig 16: Passive (left) and Active (right) Configurations (Source Gao et al [55])
Simulation results of both passive and active connections with respect to the
current waveforms under the same pulsed load conditions. These are shown in the
figure below to give a better picture of the difference in power delivery.
Fig 17: Passive (left) and Active (right) Current waveforms (Source Gao et al [55])
In the passive configuration, the battery delivered power to the load during the
pulse-on time and recharged the supercapacitor during pulse-off. The passive
system was capable of delivering a power 2.3 times greater than the battery alone
would have been able to supply. It was observed that battery current spikes were
Sヴ;ゲデキI;ノノ┞ ヴWS┌IWS H┞ デエW エ┞HヴキS ゲ┞ゲデWマ H┌デ ミラデ WミデキヴWノ┞く TエW H;デデWヴ┞げゲ ゲ;aW ノキマキデ
44
was 2.4A but peak values of almost 5A were recorded. This could be detrimental to
the battery life.
In the active configuration, a higher battery voltage was used via a step down
converter whose output was more or less the same as the supercapacitor voltage.
The converter was made to output current at a constant rate throughout the load
cycle which would either top off the supercapacitor (during pulse-off), or contribute
towards the load current (pulse-on). The deliverable peak power capacity recorded
was said to be 3 times greater than the passive configuration, and a whopping 7
times greater than the battery alone. As great as these figures sound, it came at a
cost. The total discharge time for the battery, i.e. the time it would take for the
battery to discharge from 100% SOC to the rated cut-off voltage, was 14.4 minutes
less than when the passive configuration was used. This was attributed to converter
losses. In the electric vehicle world, this could result in a decrease in range. A
comprise will have to be reached to get a right balance.
D. Shin et al [32], describes the parallel connection as an intuitive way of reducing
the load fluctuation effect on the voltage supply or dc bus. The supercapacitor in
parallel is essentially a low pass filter that prunes out rapid voltage surges; the
higher the capacitance, the better the filtering effect. To improve on this
configuration, one involving a constant current regulator that separates the battery
pack from the supercapacitor was proposed. This is shown in the figure below.
Issues associated with this arrangement include the supercapacitor charging current
control and terminal voltage control. With fixed charging current, the amount of
delivered power from the battery to the load will depend on the terminal voltage of
45
the supercapacitor which must be maintained within an appropriate or optimal
range in order to meet the load power demand.
Fig 18: Constant Current Regulator based Configuration. (Source D.Shin et al [32])
ATT R&D Co. and Ness Capacitor Co. Ltd conducted tests on a neighbourhood
electric vehicle (NEV); the Invita, to demonstrate the advantages of using
supercapacitors alongside batteries [34]. With the aid of a high power contactor, a
direct parallel connection between the battery pack (72V sealed lead acid) and
supercapacitor module (92V, 50F) was achieved. The latter electronically limited to
72V by a charge balancing circuit to match the dc bus voltage. The curb weight for
the NEV was 545Kg and a data acquisition device was setup to log vehicle speed,
battery voltage and current, supercapacitor voltage and current in real time. A fixed
15km route with light traffic congestion and frequent stop-and-go situations was
chosen to test the Invita. In each case (with and without supercapacitor), the
┗WエキIノW ┘;ゲ Sヴキ┗Wミ ┌ミデキノ デエW H;デデWヴ┞げゲ ゲデ;デW ラa Iエ;ヴェW キミSキI;デWS ; IラマヮノWデW
depletion. Results reported that on a full charge, energy consumption in Kwh/km
46
was 0.161 with supercapacitor on and 0.186 when off. This translated to a 15.4%
increase in range per charge.
Researchers in [35] have also attempted a parallel combination of a valve regulated
lead acid (VRLA) cell with a string of supercapacitors tラ キマヮヴラ┗W デエW aラヴマWヴげゲ
performance under realistic hybrid vehicle power demands. By logging voltage and
current data from a Honda insight driven on a test track, a power demand profile
was created and scaled down to represent the power profile experienced by a
single cell within the battery pack. Hawker Cyclon 8Ah cells were used along with
both EPCOS and MAXWELL supercapacitor cells. They were subjected to the same
power demand profile on a purpose built test bench over 2400 seconds. Results
show that cell current drawn was significantly reduced as the supercapacitor
sources and sinks most of the transient current. The VRLA cell acted as bulk energy
reservoir and supported the low energy storage deficiency of the supercapacitor. A
comparison of rms and maximum cell current was done with and without
supercapacitor combination. A 10% reduction was recorded with a 100F
supercapacitor, 50% reduction with 1250F and 60% reduction with 2500F as
compared to when the cell was used alone to fulfil the power demand profile.
Ohmic losses which are effectively I2R losses are the major culprit of cell heating. By
reducing the rms current from the cell i.e. heating effect, beneficial effects on cell
lifetime, electrolyte loss, gassing and corrosion are obtained as they are all
temperature dependent.
47
Control System and Energy Management Algorithms
Researchers in [33] successfully developed a prototype EV powered by a
combination of high energy sodium nickel chloride (ZEBRA) batteries (28Kwh) and
supercapacitors. ZEBRA battery was chosen due to lower costs and safety factor
even though it is known to have a low specific power density [16]. To solve this
power problem, a supercapacitor bank (20F, 300V) was installed to provide peak
power for driving the 53KW brushless dc motor, better acceleration and also
improved regenerative braking capabilities. To manage the energy flow between
the ZEBRA battery and supercapacitor bank, a buck boost type dc to dc converter
which is implementing an energy management algorithm was used. Figure X below
shows the block diagram and also the control scheme of the system. The algorithm
is based on input parameters from sensors mounted in the vehicle such as battery
voltage and current, drive current, supercapacitor voltage and current, input and
output currents of the dc to dc converter, battery state-of-charge (SOC) and vehicle
speed. The platform used is a TMS320F241 DSP from Texas Instruments which
sends corresponding gating signals to the converter. Speed provides a useful insight
into the vehiIノWげゲ WミWヴェ┞ ヴWケ┌キヴWマWミデき ノラ┘ ゲヮWWSゲ マW;ミゲ マラヴW WミWヴェ┞ ヴWケ┌キヴWS aラヴ
an impending acceleration (i.e. keep the supercapacitor fully charged) while high
speeds translate to more space for storing regenerative braking energy (i.e. empty
out the supercapacitor). Predetermined charge curves for the supercapacitor bank
based on the conditions mentioned above were also estimated using neural
networks. Results from this work show that an improvement in acceleration was
achieved as a result of the supercapacitors; 16.1% higher for 0-40 km/h
acceleration, 31.3% for 0-60km/h and 38.5% for 0-80km/h. In terms of range, a
48
10.76% improvement was recorded on a fast track and a 16.67% improvement on a
slow track. Fast track in this context refers to highway speeds with longer periods of
acceleration with little or no stops (i.e. supercapacitor has little or no chance of
recharging via regenerative braking). Slow track refers to city speeds; short bursts of
acceleration as well as braking (i.e. the supercapacitor is constantly being topped
off).
Fig 19: Block Diagram and Control Scheme of Dixon et al ZEBRA + Supercapacitor
[33]
Researchers at the University of Strathclyde developed some control strategies
based on some simple rules for a battery-supercapacitor power pack
implementation in a Shelby cobra [56]. However, only a simulation was done using
ADVISOR software which was developed by National Renewable Energy Labs
(NREL). A 78V, 70Ah gel-type lead acid battery pack and a supercapacitor series
pack of the same voltage, 83F were modelled. Two half-bridge DC to DC converters
controlled the power flow from the battery and supercapacitor respectively with a
by-pass option for when the battery solely powers the motor. This was to avoid the
49
losses associated with the converter usage. The control strategy was summarized in
the figure below.
Fig 20: Flow chart of Control developed for battery supercapacitor Shelby cobra
(Source Carter et al [56])
Two control strategies were tested over simulated drive cycles (the New European
Driving Cycle ECE); one for maximising system efficiency and the other intended to
minimise peak battery currents. The results obtained were compared with the same
simulation but using the battery as the sole power source which recorded a peak
battery current of 240A and a yield of 6.26 km/kwh. It is tabulated below:
50
Strategy Pmin Pch Peak
Battery
Current
Yield
Max. system
efficiency
6800W 0W (Supercap
only charged
during Regen.
105A (56.25%
reduction) 6.38 km/Kwh (3%
increase)
Optimised
battery life
4200W 1600W 66A (72.5%
reduction 6.11 km/Kwh (3% loss)
Table 2: Comparing the two control strategies developed by Carter et al [56].
The tabulated results showed that it is not usually possible to simultaneously
optimise the hybrid system for both efficiency and battery life; one must be
prioritized over the other. Both strategies improve the vehicles ability to meet
power demands.
The current state-of-the-art of energy management algorithms for hybrid and
electric vehicles is based on heuristics and gradually progressing towards artificial
intelligence methods such as neural networks etc. [36]. The control strategy for
battery + supercapacitor combination is largely a function of the supercapacitor
capacity [37]. Cost constraint of large super-capacitors creates the need for
advanced energy control strategies. Control strategies usually based on heuristics or
an optimization model are translated into algorithms and implemented on an
embedded system (microcontroller or digital signal processor). The embedded
system acquires all relevant signals, processes data, and generates the PWM signals
to commutate IGBTs in the DC-DC converter. Input signals include motor/vehicle
speed and power, battery voltage and current, battery and supercapacitor State-of-
charge (SOC), system temperature among others.
A heuristic energy management algorithm is based on the fact that the energy
content of the supercapacitor must have an inverse relation to vehicle speed [38].
51
Thus, when the vehicle runs at low speeds, energy is reserved to accelerate. On the
other hand, when the vehicle runs at high speeds, space to store energy from
braking is made available. This is achieved by injecting or extracting current from
the ultracapacitors to reach a predefined state-of-charge reference, which depends
on the vehicle speed. By limiting the current extracted from (or injected to) the
battery pack, its lifecycle is effectively increased. The supercapacitor array supplies
the currents whose limits are outside that of the battery pack.
Optimal control energy management algorithm is based on the application of
optimal control techniques to determine the optimal power support for a real-life
power-demand-profile [38] i.e. minimizing the energy extracted from the battery
and providing a power buffer from the supercapacitor in order to maximize system
efficiency for a particular drive cycle. Power demand data series is obtained
experimentally from driving an EV or a test bench under real driving conditions. This
data series is used as a knowledge base to train an Artificial Neural Network (ANN).
A nW┌ヴ;ノ ミWデ┘ラヴニ キゲ ; Iラマヮ┌デWヴ ;ヴIエキデWIデ┌ヴW マラSWノWS ┌ヮラミ デエW エ┌マ;ミ Hヴ;キミげゲ
interconnected system of neurons which mimics its information processing,
memory and learning processes. It imitates the brains ability to sort out patterns
and learn from trial and error, discerning and extracting the relationships that
underlie the data with which it is presented. As a result of this training, the network
;Iケ┌キヴWゲ デエW さニミラ┘ノWSェWざ ミWIWゲゲ;ヴ┞ デラ SWデWヴマキミW デエW マラゲデ Wfficient supercapacitor
current under different conditions. The optimality will depend on how many sets of
data (or the driving conditions) are used to train the network.
52
Research work [39], combines a heterogeneous energy storage system consisting of
low power; high capacity batteries and high power supercapacitors with a
ヮヴWSキIデキ┗W Iラミデヴラノ ゲ┞ゲデWマ デエ;デ ラヮデキマキ┣Wゲ ヮラ┘Wヴ aノラ┘く TエW さCエ;ヴェWC;ヴざ ヮヴラテWIデが
which authors of this work have pioneered, was aimed at achieving optimized
power management between the energy sources by utilizing information such as
vehicle state, driver history, GPS coordinates and even information available on the
web like traffic data, weather reports etc. Variables that significantly impact the
energy demands of a commuter electric vehicle (or any other vehicle) are highly
variable speeds and traffic conditions, unique individual human driving style and
elevation changes through the course of travel. Each one of these factors would
affect the design of an optimal control system for a battery-supercapacitor
combination.
The ChargeCar project has collected driving data from normal petrol volunteer
vehicles borne with GPS units and elevation data for the entire United States. This
large repository was used to test various energy management algorithms rather
than using standard drive cycles such as FUDS, UDDS etc. A robust physics-based
electric vehicle model was created which takes input GPS coordinates from the data
repository and translates it into accurate power demand for equivalent electric
vehicles. This power demand data along with vehicle, battery and supercapacitor
state was used to test out algorithms which control power flow. The physics-based
model described in this work is similar to the modelling done in chapter 3 of this
thesis where the total tractive force required by the vehicle is equal to the sum of
the acceleration force and the resistive forces due to air, rolling and gravitational
resistance. The power required by the motor to propel the vehicle was thus
53
calculated by taking into account 85% drivetrain efficiency during forward
movement and a 35% energy recovery during braking. The model for predicting the
power required by the load was verified and validated by comparing data with 2002
Toyota Rav4 EV, and using BombardiWヴげゲ ノ;ヴェW ゲI;ノW ヮWラヮノW マラ┗Wヴ ┗WエキIノWゲく
Various algorithms were tested out starting from an ideal case to predictive
algorithms. In each case, current-squared (I2) demand on the battery as an indicator
of efficiency was used as it is closely related to both internal efficiency as well as
overall battery longevity. The ideal case algorithm or optimal bound as it was
referred to, assumes, or provides the power demand profile of the particular vehicle
for the entire trip beforehand thus serving as a benchmark to subsequent predictive
algorithms. The energy management system uses rules to force the power flow to
certain levels according to the demand profile and keeping the I2
losses to the
minimum. A 70.06% reduction in current-squared was recorded which suggested a
significant amount of battery life savings; however the exact savings figure
calculation was beyond the scope of this work. Also a 15.56% increase in range was
recorded. The second algorithm tested could be referred to as a naïve buffer as it
only utilizes the supercapacitor when necessary; for acceleration and it is recharged
by regenerative braking only. There is no interaction between the battery and
supercapacitor. In this scenario if the supercapacitor is completely depleted during
a discharge or acceleration, the battery takes over the remaining responsibility. The
results of this experiment were a 9.05% reduction in current-squared on the
battery, and the 6.27% increase in range. This provided some insight as to what an
さキミデWノノキェWミデざ WミWヴェ┞ マanagement algorithm can achieve given the degree of
54
freedom of approximately 9% to 70% in current-squared and 6% to 15% increase in
range.
An improvement to the naïve buffer was developed and it was aimed at charging
the supercapacitor at a constant rate whenever the load demand was low; during
idle or stop conditions, to a maximum charge limit. This limit could be shifted based
on the speed of the vehicle. For example, based on the present speed, leave enough
room in the supercapacitor to store regenerative braking energy should the vehicle
come to a stop. At the same time, let the energy present in the supercapacitor be
sufficient for an acceleration, while at complete stop conditions, charge the
supercapacitor fully for an impending acceleration. The algorithm was built around
these rules and the results obtained came much closer to the optimal bound;
40.15% reduction in I2 and 9.55% increase in battery pack range.
In a bid to utilize the knowledge obtained from maps, routes, traffic signals and GPS
coordinates, machine learning techniques were incorporated into an algorithm
which uses the K-Nearest Neighbour (KNN) search to predict the upcoming duty of
the vehicle. At each time-step in a trip, the algorithm searches for k, in this case,7,
nearest, most simキノ;ヴ S;デ; ヮラキミデゲ キミ ; ヮ;ヴデキI┌ノ;ヴ Sヴキ┗Wヴげゲ エキゲデラヴ┞ ┘エキIエ Iラミゲキゲデゲ ラa
55hours of driving data for each driver. The nearest-neighbour search finds the
most similar neighbours and then averages these upcoming duty curves with each
other at each time slice into the future, discretized to one second. This yields an
;┗Wヴ;ェW a┌デ┌ヴW S┌デ┞ aラヴ ゲW┗Wミ ヮラキミデゲ キミ ; Sヴキ┗Wヴげゲ エキゲデラヴ┞ マラゲデ ゲキマキノ;ヴ デラ エキゲ I┌ヴヴWミデ
state. This is used as the predicted duty for the energy management algorithm,
which then simply assumes that this upcoming duty is absolute truth and calculates
55
the same brute force solution as in the optimal bound. Therefore, the gap between
the optimal bound and the performance of this algorithm hinges simply on the
accuracy of the predictions. The k-Nearest Neighbor search, which has very little
semantic knowledge of the problem, nevertheless yields decent results: a 57.69%
reduction in battery current-squared, and a 12.53% increase in range.
EノWIデヴキI ┗WエキIノWゲ I;ミミラデ IラマヮWデW ┘キデエ ICEげゲ ┞Wデ キミ デWヴマゲ ラa Sヴキ┗キミェ ヴ;ミェW. An
intelligent EMS needs to be adopted to maximize the utilization of the energy
source. According to [40], making use of sensory inputs like current, voltage, speed
as well as external factors like climate and environment, the EMS can realize many
functions. Optimize system energy flow, predict the remaining energy and hence
the residual driving range, suggest more efficient driving behavior, more efficient
regenerative braking among others. In order to increase the driving range, multiple
energy sources may be adopted for modern EVs. The corresponding combination
and hybridization ratio should be optimized on the basis of the vehicle performance
and cost.
The current state-of-the-art of energy management algorithms for hybrid and
electric vehicles is based on heuristics and progressing rather rapidly towards
artificial intelligence methods such as neural networks etc [38]. In this work both a
heuristic based algorithm and optimal control technique applied to neural
networks were implemented on a battery に ultracapacitor hybrid energy storage
ゲ┞ゲデWマく TエW ┗WエキIノWげゲ ┞キWノS ふニマっニWエぶ ┘;ゲ キミIヴW;ゲWS キミ ヵくヲХ ;ミS ΒくΓХ ┘キデエ デエW aキヴゲデ
and second algorithms, respectively. An economic analysis of costs only was carried
out and it showed that the present high cost of ultracapacitor energy storage
56
systems can only be compensated for, if it contributes to a 50% or more extension
of battery life.
Even though the cost of ultracapacitor cells are projected to fall in coming years,
the challenge of this research project would be to justify the incorporation of
ultracapacitors in an electric vehicle by showing at least a 50% increase in battery
life.
Real time control strategies and energy management system requirements of
battery supercapacitor combination proposed in [37] include a current and speed
restrained control strategy. The former ensures that the current drawn from the
main battery pack is within a certain safe limit and extra current is drawn from the
supercapacitor while the latter control strategy is based on the speed of the vehicle
i.e. at high speed, the supercapacitor should be discharged and prepared for
imminent regenerative braking while at low speeds the supercapacitor should be
fully charged in anticipation of acceleration. The results of implementing the control
strategies showed an increase in maximum speed of 68km/h without
supercapacitor to 96km/h with supercapacitor. An important observation from this
work is that the average energy consumption of the vehicle with auxiliary
supercapacitor source is reduced in city driving conditions due to constant stop-
and-go regenerative braking while in highway driving, the energy consumption does
not reduce as the speed is fairly constant.
Machine learning algorithm combined with fuzzy logic for online power
management of a multi powered source HEV based on minimizing the power loss
from each source and placing system constraints to ensure good quality power
57
service [41]. This algorithm learns about the optimal combination of power sources
(e.g. battery, fuel cell, supercapacitor etc) best suited for a given driver load request
(drive cycle) and a fuzzy controller acts on this knowledge by delivering power as
soon as the load request is made, bearing in mind all the system constraints like
state of charge of battery, ultracapacitor etc. Each power source combination is
associated with a power loss function at any time t during the load request.
According to Y.L Murphey et al [42], driving patterns exhibited in a real world driver
are the product of the instantaneous decisions of the driver to cope with the driving
environment (road type, traffic congestion, driving style, vehicle operation mode). A
neural network model was developed for predicting online driving patterns in the
near future based on the short term history of the driver during a trip. A quasi
optimal set of 14 statistical features (after a rigorous feature selection process
involving the facility specific driving cycles) was used to characterize the speed
profile of a vehicle (see table below).
58
Fig 21: 14 features selected for roadway prediction (Source Y.L. Murphey [42])
To realize real-time control, the historical data should be acquired online in a
certain time window. If the time window is too short, the historical data may not
reflect the driving cycle correctly; if the time window is too large, the computational
burden may be too heavy for real-time control. An optimal window size and time
step was proposed to divide the speed profile into segments. The figure 22 below
illustrates a segmented speed profile of the UDDS (Urban Dynamometer Driving
“IエWS┌ノWぶ ┘キデエ ; ┘キミSラ┘ ゲキ┣W ふлWぶ ラa ヱヵヰゲ ;ミS ; デキマW ゲデWヮ ふлデぶ ラa ヱヰヰゲ ふキミ ヴW;ノキデ┞ лデ
is much smaller).
59
Fig. 22: UDDS cycle segmented by a window size and a time step. Source Y.L
Murphey [42]
11 standard drive cycles were segmented as describes above for use as training and
test data. PSAT©; a simulation software was used to simulate the drive cycles. Each
segment contained a vector of 14 features which were randomly sampled into a
training and test data sets. A multi layered, multiclass neural network of 14 input
nodes, 11 output nodes with a hidden layer of 20 nodes was trained for the
roadway type prediction. An intelligent vehicle power controller that incorporates
the knowledge about the roadway type and traffic congestion level predicted by the
neural network is used to make power management decisions in a conventional car.
Significant reduction in fuel consumption was obtained even though an offline
dynamic programming algorithm was used to optimize the power management
strategy.
Y.L Murphey et al [64] also did some work on driver style classification using jerk
analysis. Jerk may be defined as the rate of change of acceleration or deceleration.
60
Mathematically, it is calculated as the derivative of acceleration/deceleration or the
ゲWIラミS SWヴキ┗;デキ┗W ラa デエW ┗WノラIキデ┞く A デヴ┌W キミSキI;デキラミ ラa ; Sヴキ┗Wヴげゲ ;ェェヴWゲゲキ┗WミWゲゲ
(which is directly related to power demand in electric vehicles) is its jerk profile
rather than its acceleration profile. Appropriate feature extraction of jerk data can
HW ┌デキノキゲWS デラ SW┗Wノラヮ ; ヴラH┌ゲデ ;ノェラヴキデエマ デラ Iノ;ゲゲキa┞ Sヴキ┗Wヴゲげ ゲデ┞ノWく F┌Wノ ヴ;デW キミ
conventional vehicles was compared with its respective jerk profile using software
simulation packages; variation in jerk results in change in fuel rate. Mathematically,
if a drive cycle is represented by DC (t), the jerk function J (t) is given as:
経系岫建岻 噺 穴態経系岫建岻穴建態
A jerk feature was defined as:
紘 噺 鯨┻経徴蛍 違
Where 鯨┻ 経徴 , is the standard deviation of the jerk over certain specified window
size and 蛍 違 is the mean jerk of the current road-type which the driver is on at that
moment. The ratio mean used is statistically called the coefficient of variation and it
is often used to depict the amount of variance between populations with different
average values. In this case different road types have different average jerk values.
Courtesy of sierra research [65], [66], 11 standard drive cycles were developed
depicting driving style on different roadway types and traffic congestion levels. The
11 cycles are divided into 4 categories; freeway, freeway ramp, arterial or urban
61
road and local roadways. A qualitative measure called the level of service (LOS) was
introduced to further classify the freeway and arterial categories based on speed
and travel time, freedom to manoeuvre, traffic interruptions, comfort and
convenience. 蛍 違 was calculated for each of the 11 drive cycles and used in a driver
style classification algorithm on a window by window basis. The algorithm is
described in the figure 22 with the thresholds for normal and aggressive being 0.5
and 1.0 respectively.
Fig 23: Driver Style (DS) Classification Algorithm Proposed by Y.L. Murphey et al [64]
TエW ゲキェミキaキI;ミIW ラa SWデWヴマキミキミェ ; Sヴキ┗Wヴげゲ ノW┗Wノ ラa ;ェェヴWゲゲキ┗WミWゲゲ ラヴ I;ノマミWゲゲ H;ゲWS
on jerk feature may be used as a key feature in energy management algorithms for
electric vehicles. Choosing the appropriate window size allows for accurate capture
of transient behaviours or jerks. If the window size is too small, it may not be
蛍岫建岻┸ 建 樺 岷岫建頂 伐 拳岻┸ 建頂峅 Calculate Jerk Profile within
a fixed window size
Standard
deviation of jerk. 鯨┻ 経徴
Determine roadway type
and current traffic
congestion level to find 蛍 違 紘 噺 鯨┻ 経徴蛍 違
Jerk Ratio
Velocity = 0; No speed
紘< Threshold (normal); Calm
Driving
Threshold (normal) < 紘 <
Threshold (aggressive);
Normal Driving
紘> Threshold (aggressive);
Aggressive Driving
62
enough time to capture the entire acceleration or deceleration event. On the other
hand, if the window size is too big, it may capture multiple events which would lead
to misclassification.
Langari and Won [43] proposed a conceptual intelligent energy management agent
(IEMA) to enable a hybrid vehicle to be driven in an economical and environmental
aヴキWミSノ┞ ┘;┞ ┘エキノW ゲデキノノ ゲ;デキゲa┞キミェ デエW Sヴキ┗Wヴげゲ ヮラ┘Wヴ ヴWケ┌キヴWマWミデ ;ミS ;ノゲラ Iエ;ヴェW
sustenance over the entire range of driving. This system incorporates a number of
subsystems such as a driving information extractor (DIE) of driving patterns for
characterizing the driving situation as well as developing a knowledge base for
torque distribution. The driving situation identifier (DSII) consists of the roadway
type identifier (RTI), driver style identifier (DSI), driving trend identifier (DTI) and
driving mode identifier (DMI). Fuzzy torque distributor (FTD) effectively splits
torque between the motor and the engine. A state-of-charge compensator was
implemented to address the issue of charge sustenance over a prescribed range.
The figure 24 below extractWS aヴラマ Jく“ Wラミげゲ PエD TエWゲキゲ ぷヴヴ], gives a graphic
explanation of the architecture of the IEMA.
63
Fig. 24: Architecture of the Intelligent Energy Management Agent (IEMA). Source J.S
Won Thesis [44]
An interesting algorithm used in this work is the Learning Vector Quantization (LVQ)
network; a prototype based supervised classification algorithm for roadway type
classification. Compared to other neural networks, the main feature of the LVQ
network is to classify input vectors into target class through a competitive layer for
identifying subclasses of input vectors and a linear layer for combining them into
the target classes. The LVQ network was trained using the statistics from 9 facility
specific driving cycles [65] to recognize driving patterns. Results were able to show
that most of the patterns were correctly classified and even the misclassified ones
were classified as the neighbour class or one with similar statistics.
The DTI assesses short term/transient features of driving cycles such as low speed
cruise, high speed cruise, and acceleration/deceleration. These features were
described by magnitudes of the average speed and acceleration values. DMI
64
classifies the drivers operating mode into startup, acceleration, cruise, deceleration
and stop or idling. FTD takes input from RTI, DTI, DMI and the current state of
charge of the energy storage device to compute an appropriate engine torque
command. Also a fuzzy driver style identifier was implemented to study the effect
of driver variability (calm, normal and aggressive driving) in the IEMA; average
acceleration and ratio of standard deviation to average acceleration as inputs and a
weight factor which compensates the torque distributor as output. A simulation
study was carried out on a typical parallel hybrid power train to evaluate the
performance of the IEMA and it was concluded that the length of window time for
each of the subsystems was critical for proper classification.
A multiple input power electronic converter (MIPEC) for interconnection among
three power sources; fuel cell (FC), battery (BT) and ultra-capacitor (UC) to power a
road electric vehicle was proposed by [45]. The control strategy of the MIPEC
includes linear control of single bidirectional boost subconverters for each source
and a supervisory system that coordinates the power flow among the power
sources to the load. The 3 sources are all below the dc-link voltage of 320V dc; FC at
150V, BT at 144V and UC at 116V. An efficiency analysis was carried out to
determine the desired operating region of each device. The converter for FC and BT
limits sudden/fast variations in charge/discharge currents. The energy management
and fuzzy supervisory system proposed by the authors of this work is shown in the
figure 25:
65
Fig. 25: Block diagram of Fuzzy Logic Supervisory Controller by A.A Ferreira et al [45]
The fuzzy logic controller has three input variables (load current, battery energy and
ultracapacitor energy) and two output variables (output correction term and fuel
cell current reference). The rule base specification depends on the designer (expert)
knowledge about the power sources and traction constraints. The rules developed
┘WヴW H;ゲWS ラミ ゲ┌IIWゲゲキ┗W W┝ヮWヴキマWミデゲ デラ Wミゲ┌ヴW デエW ヮヴラIWゲゲげゲ ヴラH┌ゲデミWゲゲ ;ミS
reliability. A summary of the rule based is described below as this will give an insight
to control algorithms which may be proposed by this thesis in future chapters:
66
Fig. 26: Fuzzy Membership Functions and Rule Base by A.A Ferreira et al [45]
Detailed information regarding the fuzzy rules can be obtained from [45]. It was
acknowledged that fuzzy logic does not guarantee optimal results in all situations,
but provides a satisfactory solution to control the overall system, allowing easy
modifications for adding on more rules to improve the performance.
An interesting work was presented by Jian Yang et al [46]; the estimation of the
driving load of a HEV by a predicition model using discrete cosine transform (DCT)
and support vector machines (SVM). The DCT was used to compress the data for
reduction of dimensionality and filtering noise. It reduced the dimensionality of the
drive cycle (historical part) data from 150 to 9 in this work. This not only
67
significantly reduced computational complexity, but also reduced the standard
deviation and slightly improved the prediction accuracy. DCT uses fewer bases to
transform the data than the FFT. Again Fuzzy logic approach was used to determine
the average driving load levels during the forecasting period. This time the input
membership functions were the vehicle speed and acceleration while the output is
the driving load. In the step of de-fuzzification, centre of gravity method is adopted
to convert the fuzzy value into numerical value.
The SVM with a radial basis function (RBF) kernel was adopted to classify the
driving load sequence into five levels of driving load (VL=very low, L=low,
M=medium, H=high, VH=very high) defined by fuzzy logic. LIBSVM [77], was utilized
as an off-the-shelf software package for the implementation. To optimize the
parameters of the SVM, a stochastic global optimization algorithm named CPSO was
adopted. The SVM was compared with variants of Neural Networks (NN) with 3-
mixed training and testing datasets from standard driving cycles in the real world
and it gives better prediction accuracy and lower computational complexity than
NN. The SVM gets better generalization ability than the cascaded neural network
with node decoupled extended kalman filtering (CNN-NDEKF), can be trained faster
than NN. The block diagram describing the prediction model is shown in the figure
27.
68
Fig. 27: Block diagram of driving load forecasting of HEV. Source Jian Yang et al [46]
In a related work, Xi Huang et al [47] and [78] attempted at correctly discriminating
driving condition which was described as an important premise to an optimal
control strategy for HEVs. The driving condition of hybrid vehicle is a complicated
variable dependent on numerous external factors such as wetness of road, terrain,
traffic congestion and weather and these factors may not have exact numerical
quantities. Thus, classifying driving conditions based on vague data is a key research
issue. Feature extraction of randomly truncated standard driving cycles was carried
out and statistically significant features such as average speed, average
acceleration, and maximum speed were chosen. Feature selection was carried out
using principal component analysis. Analysis of variance (ANOVA) was used to test
the significance of the features selected. A classification was carried out to compare
the performance of the multilayer neural network, linear classifier, k-nearest
neighbour and the SVM. The results show that the SVM is fit for classification of
linear non-separable samples giving the best accuracy of 97%.
69
Similarly, the SVM via the support vector regression (SVR) was used to estimate the
SOC of a high capacity lithium ion manganese phosphate battery pack as a function
of cell voltage, cell current and cell temperature [67]. Calculating and controlling the
SOC for EV and HEV applications is critical in order to give users an indication of
available runtime left or range, prevent detrimental over charging and over
discharging which usually leads to reduction in battery life. Due to the
unpredictability of both battery and user behaviour (driving style or pattern),
accurate estimation of soc is not straightforward. Several techniques used include
current integration (ampere/coulomb counting), artificial neural networks, fuzzy
logic, kalman filter based calculators etc. are employed. Accuracy varies in each of
these methods but the cost of higher accuracy comes with higher implementation
cost due to high computational complexity. For example, in [74], [75] and [76], the
extended kalman filter (EKF) soc estimator designed was able to achieve a ±5%
accuracy compared to the ±15% of coulomb counting method. The dual filter
tweaks the system so that the effects of battery aging and degradation are
accounted for and incorporated into the battery model in real time. However, a full
implementation of this scheme requires a 40 MHz, 32-bit floating point processor.
TエW “VMげゲ ニWヴミWノ ;ヮヮヴラ;Iエ ヮヴラ┗キSWゲ ; マラS┌ノ;ヴ aヴ;マW┘ラヴニ ┘エキIエ I;ミ HW ;ヮヮノキWS デラ
various domains such as biomedicine, bioinformatics, image analysis, machine
vision and non-linear function estimation ([68], [69], [70], and [71]) to solve
classification and regression problems. A kernel is the bridge between software
applications and the real data processing done at hardware level. Apart from [67],
[72] and [73] also used the SVM to model the EV battery behaviour. [72] Proposed
an SOC estimator for a lithium ion polymer battery which extracts support vectors
70
from a battery operation history then uses only these support vectors to estimate
the SOC, resulting in minimal computation load and suitable for real-time
embedded system applications. Basically thousands of training data from real
battery packs are taken in by the SVM and condensed into a smaller set of support
vectors which can then be manipulated by an inexpensive 8-bit microprocessor to
estimate the SOC. The experimental procedure carried out by [72] is summarised
below:
Training
Training data should be different from test data.
Training data should cover the spread of operation of the battery under
normal working conditions e.g. from 20% SOC to 85% SOC.
Data should be realistic and represent a continuous flow of measured data.
Data scaling/pre-processing is key to convergence; all vector elements in the
range of 0.0~1.0
Optimizing SVM Parameters
Cost parameter (C) controls the trade-off between allowing training errors
and forcing rigid margins of classification.
K; type of kernel function used (linear, polynomial, radial basis and sigmoid).
Polynomial kernel was used for this research.
粥 ; error constant
71
“VM ヮ;ヴ;マWデWヴゲ ┌ゲWS ┘WヴWぎ CЭ ヱンくΒヶが 粥Э ヰくヰヰヰヱが ゲЭ Αくン ;ミS ヴ Э ヱΓくΑく TエW
last two are constants from the second degree polynomial kernel K(a,b)= s x
(a.b)2+r .
Testing Data and Results
Data from simple SOC tests and dynamic SOC tests are done to obtain robust
results.
Optimal SVM tested with simple step input drive cycle; root-mean squared
error is 5%, max. +ve error is 16% and max. にve error is -9%.
Optimal SVM tested with US06 driving cycle; root-mean squared error is
5.76%, max. +ve error is +12% and max. にve error is -2%.
A root-mean squared error closer to zero means that the model is a better
predicitor.
The optimized SVM was implemented on an inexpensive 8-bit PIC18F8720
microcontroller running at 20Mhz and the SOC was computed in about 54ms. In a
related work, Wang Junping et al [73] produced a research which models an 80Ah
NiMH battery pack in order to estimate its SOC. The least square (LS) SVM algorithm
was used for non-linear dynamic estimation of battery SOC. The model has 3 inputs
(charge/discharge current, temperature and SOC) and one output (load voltage). In
battery testing, the FUDS test cycle was used to charge and discharge the NiMH
battery pack with the capacity of 80 Ah. Using a training set of 3000 data pairs, the
SVM model with the RBF kernel was obtained. Testing was carried out with 6500
data pairs and the maximum relative error recorded was 3.61%. This shows that the
72
SVM model can simulate battery dynamics reasonably well even with small
amounts of experimental data.
Variation in vehicle speed and acceleration represents driving behaviour which in
turn determines the energy required to propel a vehicle [57]. In real life, driving
behaviour is always constrained by traffic and road conditions, hence methods
capable of modelling various generic driving behaviours are critical to trip oriented
energy management systems.
Aミ ラヮデキマ;ノ ヮラ┘Wヴ マ;ミ;ェWマWミデ ゲ┞ゲデWマ ふPM“ぶ ┘;ゲ ヮヴラヮラゲWS aラヴ HEVげゲ H┞ ┌デキノキ┣キミェ
the knowledge of future terrain and traffic conditions which will make more
judicious use of the available electric energy [58]. While HEV experts acknowledge
that such information can increase fuel or energy efficiency in vehicles, it has not
been implemented yet. The figure below provides an overview of a predictive
energy management system based on 3D terrain maps.
Fig. 28: Forecasting energy usage in HEV based on 3D terrain maps (Source Chen
Zhang et al [58])
73
Rui Wang et al [59] identified three main techniques that were used to classify and
predict future driving conditions. These are by using GPS or intelligent transport
systems (ITS), statistical clustering analysis method and Markov chain based on
stochastic process prediction technique. If the origin and destination of a trip is
known a priori, GPS could be a useful tool for providing terrain information as
described in [58]. In a series HEV, GPS data was used to predict the approximate
driving distance, upcoming terrain data along with an estimated velocity [60]. Using
a rule based strategy, 13% reduction in fuel consumption as well as 8% cost
reduction was achieved. Statistic and clustering based classification involves
collecting certain historical and current driving cycle parameters in order to predict
the future pattern, typically 1-2 minutes ahead. Several literatures have identified
various driving parameters or statistic features which affect the rate of fuel
consumption and emissions. [61] Identified 10 parameters while [62] pointed out
up to 62 parameters in total and classified them into 16 groups. On the other hand,
[63] proposed only 2 parameters (average trip speed and distance per stop) which
were used to cluster driving cycles into highway, suburban and urban respectively.
The issue here is not the number of parameters identified, it is rather at what cost it
comes at. Too many cause higher hardware cost and longer computational time
which makes it not feasible to implement on board vehicles or in real time; on the
other hand, too few parameters may misrepresent or misclassify driving conditions.
Recent literature on this subject matter indicates that a reduction in the number of
parameters is desired for real-time implementation and also maintaining good
prediction accuracy. Rui Wang et al [59] concluded from extensive study that,
ultimately, some parameters seem to be more prominent than others. They are
74
average speed, average acceleration/deceleration, maximum acceleration,
maximum velocity, minimum deceleration and standard deviation of acceleration.
Iミ ; ┘ラヴニ ヴWノ;デWS デラ デエキゲ ヴWヮラヴデげゲ ヮ;rticular research interests, [48] relates
battery/powertrain performance of an electric vehicle to its driving and usage
patterns. A duty cycle in this work refers to a history of power usage of a vehicle i.e.
power versus time curve while a drive pattern was used to describe a driving
condition taking into account both road condition and driving behaviour. Time
stamped trip data and charging data was collected by an automated on-board data
acquisition device in a flash memory card during operation of the Hyundai Santa-Fe
battery powered electric vehicle.
Drive cycle is broken down into a series of sequential isolated drive pulses (DP). This
paper defines the DP as an active driving period between two continuous stops in a
trip. Distribution plot of Average speed Vs Distance is used as a basis to develop
fuzzy logic membership functions and fuzzy rules to classify driving events using the
variables of average speed and distance. By classifying a driving event for each DP, a
drive cycle profile can be constructed. Driving events: STOP N GO (SnG), URBAN (U),
SUBURBAN (SU), RURAL(R), HIGHWAY (H). The fuzzy rules for classifying driving
events were manually tweaked until a proper interpretation of driving cycles among
all trips in the database was achieved. The figure below is an example of how the
fuzzy logic pattern recognition was used to categorize a drive cycle into the 5 driving
events.
75
Fig. 29: Drive cycle analysis; fuzzy classification of drive pulses (DP).Source B.Y. Liaw
et al [48].
The concept of microtrips was introduced by Shuming Shi et al [49], defined as the
speed time profile between successive stops in a drive cycle. Principal component
analysis (PCA) and fuzzy clustering was used separate real world drive data into
microtrips. Interestingly raw velocity-time data was collected using Global
Positioning Satellite (GPS) system as opposed to the traditional sensor system.
Fifteen statsistical metrics/features based on previous research were chosen for
cluster analysis. The core idea of PCA being dimension reduction as there is a
certain correlation between these statistical features such that a linear combination
could represent the majority of the whole original feature set. Principal components
of microtrips corresponding to eigenvalues with a high sum of contribution rates
were selected to represent information of all the original 15 metrics with a
reduction in dimension. Results of the fuzzy clustering produced an optimal cluster
76
number of 11. Similarity was found between driving characteristics of the same
cluster, but great difference exists among driving characteristics of different cluster.
PCA was applied to get the most valuable coefficients from the data of drive cycles
which can be used as control input variables for intelligent control strategies for
HEVs by JianFu et al [50]. Statistical software was used to lower the dimension of
the weight factors which correlate with fuel economy in HEVS. Here, 14 metrics
were used to describe driving data. Time, distance, max speed, average speed,
maximum acceleration, maximum deceleration, average acceleration, average
deceleration, idle time, number of stops, maximum up gradeability, average
gradeability, maximum down gradeability, average down gradeability. Drive cycle
data can be quite redundant and finding a correlation between metrics of the data
can be complicated. The feasibility of PCA was verified using Kaiser Meyer Olkin
KMO ;ミS B;ヴデノWデデげゲ デWst. Three principal components instead of multi-dimensional
variables were derived and would be used for an intelligent control strategy to
reduce fuel consumption in HEVs.
Chan Chiao et al [51] tackled power management in hybrid electric vehicles from a
stochastic viewpoint. A design method for the power management control
algorithm for hybrid electric vehicles was developed by using Markov chain
modeling and stochastic dynamic programming techniques. The driver power
demand was modeled as a Markov process to represent the future uncertainty of
the driver power request under diverse driving conditions. As opposed to
deterministic optimization over a given driving cycle, the stochastic approach
optimizes the control policy over a family of diverse driving patterns. The infinite-
77
horizon SDP solution generates a time-invariant state-dependent power split
strategy, which governs the engine and battery operations. The algorithm can be
directly implemented in simulations and vehicle testing. Simulation results indicate
that the SDP control strategy achieves improved performance in almost all driving
scenarios over the sub-optimal rule-based control strategy which is trained based
on deterministic DP-results. Furthermore, the proposed approach provides a
directly implementable control design path, which is highly desirable because of its
potential for a fully integrated optimal design and control process.
Genetic Algorithm (GA) was used to tune the parameters of a fuzzy logic controller
based on similar driving pattern recognition and prediction for the energy
management of a HEV [52]. The concept of micro-trips described in previous work
was used. Driving pattern was classified based on the distribution of average speed
for all collected microtrips; Congested condition (Average speed <7kph), urban
(Average speed 7-15.5kph), Extra urban (15.5 に 23.5kph) and highway (> 23kph). A
driving prediction model which contains traffic conditions of driving data history
either offline or real time was proposed in this work using hidden markov model
(HMM). In the case of an online (real time) operation, new driving experiences are
used to update the model constantly. HMM is a popular stochastic tool for studying
time series data. The control strategy was based on changing the membership
function of the fuzzy logic control in response to changes in driving condition. The
objective function to be minimized in this case was the integral of fuel consumption
and engine emissions over a drive cycle. An appropriate fitness function was chosen
to evaluate the status/fitness of each possible solution to the optimization problem.
More information about the GA parameters used can be obtained from [52]. Finally,
78
simulation results show the effectiveness of the approach where reduction in the
fuel consumption is achieved.
The concept of self learning of driving cycle was described in this work [53]. Real
time speed profile was collected by an in vehicle device and a statistical analysis was
done to calculate 28 characteristic parameters of the drive cycle. These were used
as inputs to a self organizing map (SOM) to classify into clusters of one of three
drive cycle representatives. The result of the classification was used to manipulate
the driving control strategy accordingly for optimum performance. Self learning
here means the use of a general packet radio service (GPRS) to transmit drive cycle
data and also receives control parameter updates in real time.
On a rather theoretical level, several researchers have developed energy
management and power management techniques that apply priory information
regarding the vehicle propulsion power demands. These methods do provide a
means to identify the maximum obtainable improvements in terms of energy
efficiency and performance benefits. The findings also clearly support the grounds
for further research in this area. However, in spite of significant contributions, there
have not been many attempts to address the complete implementation process of a
working system. The figure 30 sums up the literature survey.
80
CHAPTER 3: METHODOLOGY
This chapter begins by describing the equations governing the movement of an
electric vehicle on a level or inclined road surface. The propulsion system has to
overcome certain forces (which consists of various components) in order to
accelerate or retard the vehicle.
Using these equations, various software models were developed in MATLAB
Simulink © environment in order to simulate the performance of the proposed
experimental test electric vehicle conversion.
The third section of this chapter describes the step by step design and construction
of the experimental test vehicle. Mechanical effort such as engine removal, motor
coupling, battery and supercapacitor mounting are described. Also, the traction and
control wiring, sensor setup, data acquisition system, on-board real time display of
circuit parameters were laid out in detail.
The section that follows is the on-the-road test of the electric vehicle along a fixed
route in the university premises and important drive cycle data is collected and
stored for analysis. Statistical feature extraction of drive cycle data by dividing
Sヴキ┗キミェ S;デ; キミデラ ゲ┌キデ;HノW さマキIヴラデヴキヮゲざ ┘;ゲ SラミW ;ミS ;ノゲラ デエW SWゲIヴキヮデキラミ ラa Sヴキ┗キミェ
conditions based on statistical features of microtrips (i.e. urban, highway, stop and
go, idle).
81
Fig 33: An Overview of Chapter 3
EV EQUATIONS
Acceleration force
Aerodynamic drag
Rolling Resistance force
Total Traction Force & Power
SOFTWARE MODEL OF A SMALL EV
Modelling the EV equations in MATLAB ®
Simulating the total power required by a
small EV
Simulating the range & battery discharge of a
deep cycle lead acid battery
EV CONVERSION; ELECTRIC KANCIL
Mechanical Coupling
Electrical Wiring
Instrumentation & Data acquisition
DATA COLLECTION
Route specific drive cycle collection
Analysis
82
3.1 EV DESIGN CRITERIA
The design of an electric vehicle should be based on the most important task it is
required to accomplish; a high speed (performance) car, a long range vehicle or a
utility commuter vehicle midway between the two i.e. with reasonable
performance and acceptable range.
An electric vehicle should be as light in weight as possible, streamlined with its body
optimized for minimum drag, minimum rolling resistance from its tires, brakes and
steering, and optimized for minimum drive-train losses [3].
Fig 33: The forces acting on a car (Adapted from L.C Rosario [54])
Figure 33 above describes the forces that a vehicle is subjected to as it tries to
accelerate or decelerates. These forces consist of several components and are
described below:
83
Acceleration force (繋挑銚)
Tエキゲ aラヴIW キゲ SWヴキ┗WS aヴラマ NW┘デラミげゲ ゲWIラミS ノ;┘ ラa マラデキラミく
繋挑銚 噺 兼 穴穴建 撃掴痛岫穴建岻 噺 兼欠
Where a is the linear acceleration of the point mass m travelling at a varying
tangential velocity撃掴痛. Weight (mg)
Weight affects acceleration, climbing, range and speed of the vehicle. 繋直掴脹 is the
gravitational force acting on the vehicle on non-horizontal roads. When the vehicle
is on a level road, this force is very close to zero.
F gxT mgsin
An uphill acceleration of the vehicle results in a positive force while a downhill
motion results in a negative force. is the angle of inclination
Aerodynamic drag of a vehicle 繋凋帖
This part of the force is due to the friction of the vehicle body moving through the
air. It is a function of the frontal area, shape, protrusions such as side mirrors, ducts
and air passages, spoilers, and many other factors. Vehicles with internal
IラマH┌ゲデキラミ WミェキミWゲ SキSミげデ ミWWS to be aerodynamically perfect as they packed
enough horsepower at their disposal. However, batteries only provide about one
84
percent as much power per weight as gasoline and as such EVs need to be more
aerodynamically aligned. The aerodynamic drag force can be expressed as:
繋凋帖 噺 なに ┻ 貢┻ 系鳥┻畦┻ 撃捲建態
A is the vehicle equivalent frontal area in m2; the effective area a vehicle presents
to the onrushing air stream. 系鳥┻ is the coefficient of drag and it has to do with
streamlining and air turbulence flows around the vehicle. The coefficient of drag is a
measure of how easily a vehicle slides through the air. These characteristics are
inherent in the shape and design of a vehicle. Typical values for 系鳥 include: Cars
0.30 に 0.35, Vans: 0.33-0.35, Trucks: 0.42-0.46. It should be noted that modern
vehicles may have lower or even higher coefficients, for example the Toyota Prius
has a 系鳥 of 0.26. Lowering the coefficient of drag is very desirable even in internal
combustion engines as it helps to save fuel. An optimized electric vehicle with a
very low 系鳥 should have a finely sculptured rear, covered wheels, thin tires,
enclosed underbody, and low nose with sloping windshields. 撃捲建 is the vehicles
speed ; since drag depends on velocity squared, increasing speed can have a
dramatic negative effect. 貢 is the air density in kg/m3, a typical value would be 1.25
kg/m3.
85
Rolling Resistance (繋追墜鎮鎮) The rolling resistance is primarily related to the tires of the vehicle and secondarily
related to the bearings and gears. It is a constant usually approximated and is
independent of the vehicle speed. The rolling resistance force (繋追墜鎮鎮) is given as: 繋追墜鎮鎮 噺 航追追 ┻ 兼訣
Where 航追追 is the coefficient of rolling resistance. The main factors controlling 航追追
are the type of tyre and the tyre pressure. 航追追 is estimated at 0.015 on a hard
surface (concrete), 0.08 on a medium hard surface and 0.30 on a soft surface (sand)
[Bob Brant]. Tires support the electric vehicle and battery weight while cushioning
against shocks; develops front-to-rear forces for acceleration and braking and
develop side-to-side forces for cornering. Ideally EV tires should be thin giving little
contact area with the road, hard, and large diameter (higher mileage).
Total Tractive force 繋脹眺
This is the minimum force required to move a vehicle from standstill and it is given
as: 繋脹眺 噺 繋鎮銚 髪 繋直掴脹 髪 繋追墜鎮鎮 髪 繋凋帖
86
The instantaneous traction power 鶏脹眺 can be expressed as: 鶏脹眺岫建岻 噺 繋脹眺岫建岻┻ 撃掴痛岫建岻
If P TR > 0 then the vehicle is in traction mode with positive traction
If P TR < 0 then the vehicle is in braking mode with negative traction
If P TR = 0 then either vehicle is coast mode (resistive forces equal to change in
kinetic energy) or the vehicle is at rest.
87
3.2 SOFTWARE MODELLING OF A SMALL ELECTRIC VEHICLE
Based on the equations described in the previous section, a custom routine was
written in MATLAB® m-file to estimate the instantaneous traction power of a small
electric vehicle (see appendix for detailed matlab code). The input to this model is
the New York drive cycle (NYCC) which features a mainly low speed, idle and stop-
go condition. This drive cycle is ideal for this research work because the electric
vehicle to be developed in the next section will be tested on a similar route. The
outcome of this model is important because it serves as a yardstick for sizing the
electric motor and batteries which will be used in order to meet the traction power
demands of the vehicle.
The parameters for a typical small EV are assumed as follows:
Cd (coefficient of drag) = 0.19 A (frontal area) = 1.8m2
Urr (rolling resis. Coeff) = 0.015 Mass of Vehicle + 2 Passengers = 820kg
Tire radius = 0.261m 155/70R12 ng (estimated gear efficiency) = 0.91
Density of Air = 1.25 kgm-3
Road gradient = 00
In order to account for losses in transmission, gears, electric motor, accessory
devices, etc, a total system efficiency of 91% was used to simulate the model. From
the figure below, a peak power of 17.08KW was recorded and an average non idling
power of 7.23KW. This means that a motor with an average power rating of
between 8KW and 10KW with a peak of up to 20KW would be suitable to propel
this vehicle in question.
89
The power required to make the vehicle move at a certain speed as described by a
drive cycle in the figure above, directly relates to electrical power required by the
electric motor from the battery pack. A deep cycle lead acid battery pack is
simulated by coding the equations to calculate its open circuit voltage, state of
charge and depth of discharge in matlab which have been described in chapter 2 of
this work. Also the current which is drawn from the battery when it is operating at a
certain power was simulated (see appendix for code).
Fig 35: Simulating the Depth of Discharge of a Lead acid Battery pack versus
distance travelled (range)
90
Fig 36: Simulating the Current drawn from the battery pack @ 48V 207Ah
Usually deep cycle batteries are designed to be discharged from 50% to 80% of its
rated capacity [113]. They can be discharged down to 20% but this is detrimental to
the health of the battery. A healthy 50- 60% is usually desirable. From the figure 35
above, the simulated EV can do a 27km route with the depth of discharge at about
62% i.e. remaining charge is 38%; again following the New York drive schedule. The
current drawn from the battery pack reached peaks of 329.2 amperes. Assuming
instantaneous battery voltage of 48V, this roughly estimates the peak power
demand as 15.8KW and average power of 2.2KW.
The Modeling of the basic parameters or performance criteria of a small electric
vehicle using matlab® has put things into perpective regarding the conversion of a
small city electric vehicle.
91
3.3 INTEGRATION OF SUPERCAPACITORS INTO EV
This part of the project aims at proving beyond any reasonable doubt that the
integration of supercapacitor banks into electric vehicles is justified by the following
performance criteria.
(1) A significant improvement in battery life.
(2) Increase in range per charge
(3) Improved acceleration
In order to achieve the objectives above, we have to create a platform or basis on
which testing, research and development can be carried out. The testing of the
various system configurations may be done exclusively through software. The cost
is relatively low as the resources needed are computers and specific software to
carry out the simulation such as ADVISOR, PSAT, PSIM and VTB. Several researchers
have already carried out similar simulation studies [105], [106], [107], [108]. In
general, this procedure is simpler than the rest of approaches, but is also less
accurate due to the impossibility to compare it to real world measurements. Most
research simulations on EV configurations just end there without any real hardware
implementation or breakthrough.
A test bench may be setup to emulate an electric vehicle drive performance. The
test setup can be roughly divided into four main components: a dynamometer
(programmable load motor), a real time data acquisition & processing system,
power source (batteries + supercapacitor), and the electric propulsion system;
motor under test (electric motor). This could be termed as a Hardware-In-the-Loop
92
(HIL) simulation where part of the system is real hardware (batteries,
supercapacitor, dc dc converters) and part is software or artificially simulated
(electronic load to emulate real world driving conditions). The figure below shows a
sample EV test bench which may be used for this project. On-road tests sometime
are not always possible to thoroughly test each sub system. These tests are likely
done during very later phase. It is not easy to change too much configuration of the
propulsion system on a test vehicle.
Fig 37: Setup for a Sample Electric Vehicle laboratory Test Bench
For the purpose of this research work, real EV capable batteries, motor and
supercapacitor module would be purchased. However, it was discovered that a
single piece of equipment which is able to emulate the driving load of a real world
EV ふヮヴラェヴ;ママ;HノW ノラ;Sぶ ┘;ゲ HW┞ラミS デエW ヴWゲW;ヴIエ ェヴラ┌ヮげゲ H┌SェWデ ;デ デエW デキマWく Aミ
experimental test vehicle would be constructed; basically a conversion from a
93
conventional ICE to electric. It would be fitted with state of the art data acquisition
system which monitors the interaction between the energy and power sources as it
tries to fulfil the load requirements of the vehicle. In this case, the load on the
propulsion system comes from 100% real world driving. The only downside
envisaged from this setup is the inability to replicate the exact same driving pattern
over a fixed route.
As a preliminary work, an electric bicycle to be powered by a battery -
supercapacitor hybrid combination was implemented. This is a scaled-down version
of the actual electric vehicle to be implemented. Details of this work can be found
in appendix c of this report.
94
Fig 38: Work flow for EV conversion
EV CONVERSION
Mechanical mounting (adaptor plate, battery compartment, controller,
brake system modification etc.)
Electrical Wiring (controller, battery, supercapacitor, charger circuits,
dc dc converters etc)
DATA ACQUISITION SYSTEM
Installing current, voltage sensors for battery and supercapacitor
GPS antenna for logging real time speed and elevation data
Datalogger module which collects and stores data automatically
DATA COLLECTION
Collection of driving data over a fixed route
OFFLINE DATA ANALYSIS
Statistical analysis of driving data
Determining significant drive cycle features
Classifying drive cycles into different driving loads
Quantifying the effect of the supercapacitor integration
95
3.3.1 Experimental Test Vehicle
Being a pioneer researcher on electric vehicles at the University of Nottingham
Malaysia campus, generating interest or general awareness about electric cars
among undergraduate students as well as faculty members becomes important in
order to build a good research team for the future. Hence the decision to build an
actual electric vehicle by converting a regular compact ICE city car was reached.
A considerable effort of this research work has been dedicated towards the
development of an experimental vehicle to serve as a setup for investigating the
effects of supercapacitor integration into electric vehicles. This section describes
the experimental vehicle and the corresponding energy storage systems; deep cycle
lead acid batteries and supercapacitor bank. Starting form a pure battery driven
vehicle, the energy system was then augmented with the addition of a
supercapacitor H;ミニく Aヮ;ヴデ aヴラマ マ;ミ┌a;Iデ┌ヴWヴげゲ キミaラヴマ;デキラミが デエWヴW キゲ ┗Wヴ┞ ノキmited
experimental data on supercapacitor field testing. As such, the vehicle provided a
means to obtain unbiased empirical data to substantiate research claims and also to
serve as a test platform for further work. Most of all, it provided a hands on
experience.
A famous city car in Malaysia, the Perodua Kancil, was chosen for conversion into a
fully electric vehicle due to its light weight, readily available spare parts and also
suitability for a lower voltage conversion. It has a 660 cc (1997 model), three
cylinder carbureted engine rated at 31 Hp (22.1KW), other specifications can be
found below:
96
length 3365mm Engine 659cc, water cooled , 4
stroke, in-line 3 cylinder
Height 1405mm Max Output
power
22.8 KW (31hp) / 6400rpm
Wheel base 2280mm Max Torque 49Nm /3200 rpm
Kerb weight 681 kg Fuel system Carburetor; 32 liter fuel tank
Seating 5 Power train Clutch, 5-speed Manual
Coeff. Of Drag 0.37 Tires 155/70R12
Frontal Area 1.69m2 Gear ratio
1st
: 3.500 , 2nd
: 2.111 , 3rd
:
1.392, 4th
: 0.971, 5th
: 0.794,
Final: 4.722
Table 4: Technical Specifications of the Perodua Kancil [100]
The first stage in any EV conversion is usually very mechanical. This involves
removing the engine block totally from the vehicle which will make way for the
electric motor. Also, the fuel tank was taken out and there was a weight reduction
of about 150kg. Battery racks made of solid cast iron were fabricated and fitted to
the rear compartment of the vehicle as shown in the figure below. Eight (6V,
225Ah) Trojan T105 deep cycle flooded type lead acid batteries were connected in
series to produce a 48V, 225Ah battery pack. The overall weight of the battery pack
was 240Kg; concentrated in the rear compartment due to convenience of
installation and lack of space in the front compartment. However tougher coil
springs were used instead of the normal springs installed in the vehicle in order to
reinforce the rear suspension. The performance curves for the Trojan battery are
simulated in the curves below.
97
Fig. 39: Estimated Discharge Curve for Trojan T105 deep cycle battery
From the curve it can be deduced that for a constant discharge rate of 100A, the
battery is able to sustain the load for approximately 100 minutes. Also if it is
discharged between 40%-70% of its capacity, estimated cycle life is 800 に 1500
cycles.
Fig. 40: Estimated life cycle Curve for T105 battery
99
3.3.2 Electric Motor Mounting & Coupling
In an internal combustion engine (ICE), the clutch is used to disengage the
transmission from the engine (idle) and also to bring the vehicle up to speed in
ェW;ヴゲく Aミ WノWIデヴキI マラデラヴげゲ RPM I;ミ HW W;ゲキノ┞ ┗;ヴキWS aヴラマ ┣Wヴラ デラ マ;┝キマ┌マ ;デ a┌ノノ
torque eliminating the need for a clutch and also the flywheel which is used for
building up inertia between the power strokes of the ICE. Hence drivability is only
about shifting from second to third gears without a clutch which many cars can do
rather smoothly.
The clutch, flywheel and pressure-plate assembly was removed and the electric
motor was attached directly to the transmission input shaft by using a custom
made adapter plate. The plate itself bolts to the transmission, a spacer if required,
and a coupler or hub to connect the shaft of the motor to the transmission shaft.
One side of the coupler was attached to the electric motor by a keyed shaft while
the other end was fitted with a splined shaft to match the spline on the drive shaft
coming from the gearbox. This provided a simple yet sturdy coupling of the electric
motor to the clutch- less transmission which is capable of transmitting a large
torque .Custom-made adaptor plate was fabricated with 5/8 inch hardened steel
that mates the electric motor to the original transmission by matching the bolt
pattern of the transmission on one side, and that of the motor on the other side. It
must hold the motor shaft in exact alignment with the transmission shaft. The
original mounting points to the chassis of the car were used where possible, similar
mounting points were created using standard engine mounting rubber. The motor
100
was finally placed in with the help of a jack and then locking the mounting points in
place.
Fig 42: Electric Motor with coupler (center), custom made adaptor plate (top right)
and a spacer (bottom left).
Fig. 43: Brushless DC Motor successfully mounted, adaptor plate matches the bolt
pattern of the original transmission block. Reinforcement bracket is added to make
the mount more stable.
101
The controller was mounted very close to the front grill of the vehicle as shown in
the figure below on an aluminium heat sink plate. The vehicle will not be driven
during wet or rainy conditions hence there was no need to make the controller
waterproof.
Fig 44: 500A Motor Controller successfully mounted in the same position as the
radiator of a conventional vehicle
102
3.3.3 Electrical Wiring
After all the main mechanical conversion part was done, the next stage was to wire
it all up. The table below lists out all the parts which were used for this conversion
and their specifications.
Equipment /Part Specification
1
Golden Motors HPM-
1OkW Brushless DC
Motor
Model: HPM-10KW -- High Power BLDC
Motor
Voltage:48V/72V/96V/120V
Rated Power:8KW-20KW
Peak up to 20KW
Efficiency: 91%
Phase Resistance (Milliohm): 3.1/48V;
6.0/72V; 18.0/120V
Phase Induction(100KHZ): 34uH/48V;
77uH/72V; 252uH/120v
Speed: 2000-6000rpm (customizable)
Weight:17Kgs Casing: Aluminium
Length (height): 170mm Diameter:
206mm
2 HPC Series Brushless
DC Motor Controller
Voltage
Ranges:48V(~60V)/72V(~90V)/96V(~140
V)
103
Rated Current (Max):500A
Weight:2.9Kgs Dimensions:
192x208x77mm
Features: Programmable via USB port,
Regenerative Braking
Temperature Protection circuit
3 TROJAN T105 Deep
cycle Flooded/wet
lead-acid battery
6V ,225Ah (20hr rate)
185Ah (5hr rate)
28kg
4 MAXWELL
ultracapacitor module
BMOD0165P048
165F, 48.6V
ESR 7.1 milliOhms
13kg
Individually balanced cells
Compact, rugged ,fully enclosed
Max. continuous current 150A
5 12V Albright contactor
SW200B-84 12V
12VDC continuous duty coil with
blowouts
Weld resistant silver alloy contact tips
70% duty 300A
6 Foot Throttle 0-5V voltage output
micro switch
Casted aluminum
Water resistant
104
7 Emergency
disconnect/kill switch
Heavy duty 500A
Key operated
8 CR Magnetics Hall
Effect Current and
Voltage Transducer
0.001 precision
Input current:0-600A,output: 0- 5VDC
Input Voltage:0-60V , Output: 0-5V
Sensor Power Supply: 24V dc.
9 Panel meters: Voltage
,current (analog and
digital meters)
Voltage (0-80V)
Current (0-300A) with 50mv shunt
10 Auxiliary DC- DC
converter
Input: 40 64 VDC
Output: 13.2Vdc , 12A, 158W
Efficiency: 82%
11 Automotive Lead acid
battery charger
Input: 240Vac, 50Hz
No of batteries: 4 x 12Vdc or 8 x 6Vdc
Charging current: adjustable (20A Max)
12 Data Acquisition Card;
DataTaker DT82E
Low power design for remote
applications
12-24V dc input
18 bit resolution
FTP for automatic data transfer
12V regulated output to power sensors
Up to 6 Analog (+- 30V) sensor inputs
105
USB memory for easy data and
program transfer
Interface with smart sensors such as GPS
antenna.
13 GPS antenna; EM-406A Sirf star III chipset
Output NMEA 0183 and SirF binary
protocol
20 channel receiver
Hot start 1s; warm start 38s; cold start
42s
10m Positional accuracy
30mm x 30mm x 10.5mm, 16g
LED status indicator and battery backed
RAM
Table 5: EV conversion parts and specifications
The parts listed in the table above were connected according to a certain wiring
diagram (see figure 41). The thick lines represent 2/0 gauge welding cables which
were used for the main traction wiring. It should be noted that wherever possible,
fuses (10A, 5A) were used to protect the auxiliary circuit. The 12V chassis ground
must be isolated from the 48V system. This is to ensure that through error, or
accident, the chassis ground cannot complete a 48V potential loop. The 48V system
has no common grounded chassis like the 12V system.
106
In conventional vehicles, the 12V auxiliary battery used for startup and accessories
is usually charged by the alternator when the engine is running. In this EV
conversion, a step down dc-dc converter was used to replace the alternator for
charging the 12V battery. This converter taps 48V from the main battery pack and
steps it down to 13.2V dc. Its operation is controlled by a relay which is activated by
the ignition key switch of the car.
For safety reasons, the micro switch on the throttle was used to turn the main
contactor on and off, in other words if your foot is off the throttle, the main
contactor goes off. If something goes wrong such as a runaway motor, then by
releasing the throttle (usually happens on instinct), the main contactor is turned off
and the circuit is broken.
Also an emergency kill switch was installed in an accessible area (just beside the
gear stick). This serves as emergency disconnect from battery pack in case of
motor/controller failure.
108
3.3.4 Installation and testing of Supercapacitor Module
Custom made rack was fabricated to house the supercapacitor module in the front
compartment, right above the electric motor mounting assembly. A 300A fuse was
connected as well as digital voltmeters and ammeters in order to monitor the
charging and discharging capabilities of the supercapacitor module. Figures below
show the module before and after installation in the electric vehicle.
Fig. 46: BMOD0165P048; 165F, 48.6V supercapacitor which weighs 15kg
Fig 47: BMOD0165P048 installed just above the electric motor. Top right is the high
current contactor which is activated by a 12V coil and a manual switch on the
dashboard of the vehicle.
109
Using an online tool provided by the manufacturer [112], a constant current
discharge profile was simulated for the supercapacitor module. This is shown in the
figure below:
Fig. 48 Constant Current Discharge Profile for BMOD0165P048 Supercapacitor
Module
110
3.3.5 Brake system modification
Almost all cars these days have vacuum assisted braking system. The perodua kancil
does. Since the vacuum source; the engine, was removed, an alternative source of
vacuum is needed to restore the functionality of the power brakes. A 12V vacuum
pump which is controlled by a pressure switch was installed to act as vacuum
booster for the power brakes. The figure below shows the wiring diagram.
Fig. 49 Vacuum pump wiring diagram for brake system modification [111]
Fig 50: Electric Vacuum Pump (left) connected to the vacuum reservoir on one side
and the brake booster on the other. A pressure sensor turns the pump on and off
depending on the set level.
111
The pressure switch monitors the pressure in the vacuum tank. Once the suction
pressure drops to 400millibar, the switch turns on the pump via the relay. As soon
as 600millibar is attained, the vacuum is switched off. This way, vacuum is
automatically maintained between these two limits. The vacuum pump was
mounted on brackets firmly welded on to the chassis. Rubber mounts were used to
mount the pump in order to minimize the vibration which would otherwise be
transmitted onto the body of the vehicle.
Testing of the electric vacuum assist brakes was done and found to be satisfactory
and on par with conventional vacuum assist from the engine. The vacuum pump
turns on only after the brakes are depressed at least twice, draws a current of 3A at
12Vdc (36W) during operation and turns off when sufficient vacuum pressure is
achieved.
Fig 51: front view of the EV with all the necessary connections in place.
112
3.4 DATA ACQUISITION SYSTEM AND TESTING
Fラヴ キミキデキ;ノ デWゲデキミェ ヮ┌ヴヮラゲWゲが ; aキ┝WS Sヴキ┗キミェ ヴラ┌デW ┘キデエキミ デエW ┌ミキ┗Wヴゲキデ┞げゲ I;マヮ┌ゲ ┘;ゲ
chosen. A data logging device with external USB storage was setup in the car to
collect data from current and voltage sensors which were used to monitor the
power flow from the battery pack and the supercapacitor module respectively.
Fig 52: DataTaker DT82E Data logger with flash USB storage logs battery pack
current and voltage, supercapacitor current and voltage, GPS timestamps as well as
GPS speed.
The features of the data acquisition system include
DataTaker DT82E logger: The DT82E is a smart data logger designed
especially for outdoor monitoring. It is a robust, low power data logger
featuring USB memory stick support, 18-bit resolution, extensive
communications capabilities and built-in display [101]. It has 6 analog input
channels, 4 bi directional digital input output channels, a smart serial sensor
113
port which was connected to a GPS antenna in this project and a 12V
regulated power output.
Hall Effect current transducers and voltage transducers: CR magnetic
CR5210-600 provides a DC signal (between 0-5V) which is proportional to a
DC sensed current. Two of these units were used to monitor the current
drawn out of the battery pack and also supercapacitor module. This is
shown in the figure below; the maximum dc current that can be sensed is
600A [102]. CR5310-100 dc voltage transducers provide an output DC signal
(also 0-5V) that is proportional to the input DC voltage of the battery pack
and supercapacitor module respectively; the maximum dc input voltage
allowed is 100V [103]. The wiring diagram for these 2 sensors are described
in the circuit below; the sensors require a 24V power supply which is
provided by 2 12V sealed lead acid rechargeable batteries housed in the
glove compartment of the vehicle. The outputs for these sensors are fed
into the analog input channels of the DT82E data logger module.
Fig 53: Hall-effect current transducers used to monitor battery and supercapacitor
currents.
114
Fig 54: Differential Wiring diagram for Current and Voltage Transducer to data
logging device [102], [103]
Fig 55: Dashboard Mounted Display; Top left: Supercap Voltage, Top Right:
Supercap Current, Bottom Left: Battery Voltage, Bottom Left: Battery Current
115
GPS receiver module: The EM-406a is a 20 channel compact, high
performance, low power consumption GPS engine board (see figure 52
above). It uses the SiRF Star III low power Single GPS chipset which can track
up to 20 satellites at a time, and can perform extremely fast time to first fix
in weak signal environments. It is suitable for automotive navigation,
personal positioning, mobile phone navigation, marine navigation among
others [104].
Analog and digital panel meters which were mounted on the dashboard (as
shown in figure 55 above). The analog panels display the battery pack
voltage and current while the digital panel meter displays the
supercapacitor voltage and current.
Fig 56: Wiring diagram for GPS antenna and MAX232 level converter pin-out
to DB9 connector
116
Figure 56 above describes the connection between the GPS module and the serial
port of the data logger module. A MAX 232 chip was used to convert the TTL output
from the GPS module to RS-232 serial signal via a DB9 connector which plugs into
the data logger serial port. A custom DataTaker code which is able to activate the
GPS module, read the input strings and log important parameters such as GPS
speed, altitude and total distance travelled was written. At the same time, readings
from the current and voltage sensors are also logged following the same time
stamp as with the GPS data. A copy of the program can be found in the index of this
report. A real time display of the logged parameters was also displayed on the lcd
screen of the data logger device. This can be seen in figure 52 where a battery pack
voltage of 50.4V is displayed.
117
3.4.1 Driving Data Collection
Two fundamental methods have been used in the past to collect driving data; the
chase car method and the on-board vehicle instrumentation method [109]. As the
name suggests, the chase car protocol involves the installation of a range-finder
laser system on a chase vehicle which collects second-by-second speed-time
profiles from target vehicles. The on-board vehicle instrumentation involves
installing the test vehicle with all the necessary sensors and transducers to measure
デエW ┗WエキIノWげゲ ヴW;ノ デキマW ゲヮWWS-time profile. With the advent of GPS technology, non-
contact method of logging speed (as opposed to rpm sensors) can be implemented
for collecting on-road data. This research work adopts the on-board
instrumentation method coupled with GPS technology. A data acquisition device is
integrated into the system to store on-road data continuously. In general, drive
cycle construction methods typically include the following steps:
Collecting real world driving data
Segmenting the driving data
Constructing cycles
Evaluating and selecting the final cycle.
Depending on the type of driving activity that is being using to construct the cycle,
existing cycle construction methodologies for light-duty vehicles can be generalized
into four types: micro-trip cycle construction; trip segment-based cycle
construction, cycle construction based on pattern classification, and modal cycle
construction [110]. Traditionally, drive cycles have been used to test conventional
vehicle polluting emissions and fuel consumption.
118
Fig 57: The UDDS drive cycle and some of its features
Fig 58: The HWFET drive cycle and some of its features
119
Figures 57 and 58 above represent existing standard driving schedules for city and
highway driving in the United States. In comparison with ICE vehicles, the testing of
electric and hybrid vehicles pose some unique problems that are specific to the
デWIエミラノラェ┞ ┌ゲWSく Fラヴ W┝;マヮノWが EVげゲ I;ミ HW SWゲキェミWS aラヴ ヴラ┌デW-specific applications
i.e. low speed city driving or specific neighborhoods only, therefore they will
perform badly when subjected to other types of drive cycle. Standard driving cycles
for ICE-vehicles are not necessarily adequate for vehicles with electric drive trains
and provide insufficient basis for the comparison of different driveline technologies.
For power sources such as batteries in electric and hybrid vehicles, assessments on
their performance are, most of the time, conducted in laboratories on test bench
setups. Similar to standard driving schedule tests and analyses, these laboratory
tests and duty cycle analyses have constraints in their validity to real-life operation.
A main issue exists in both cases due to the problem with real-life operation where,
even under specific driving cycles or duty cycles, energy consumption strongly
depends on ambient operating conditions that are typically uncontrolled [48].
It is with this mindset that we decided to create our own driving schedule called the
UNMC cycle. Due to existing local traffic regulations, an electric converted vehicle is
ミラデ ヴラ;S ノWェ;ノが ゲラ デWゲデキミェ ┘;ゲ SラミW ┘キデエキミ デエW ┌ミキ┗Wヴゲキデ┞げゲ campus. The route
chosen is 1.04 ニマ aヴラマ デエW ┌ミキ┗Wヴゲキデ┞げゲ WミェキミWWヴキミェ a;I┌ノデ┞ ふヮラキミデ A キミ aキェ ヱΓぶ
through point C (cafeteria), around the accommodation complex and back to the
original starting point. This route is very similar to city driving conditions or speed
restricted driving conditions such as in residential areas with a lot of speed bumps.
We shall refer to this drive route (cycle) as UNMC cycle.
120
Fig 59: Fixed Driving route for EV testing at the University of Nottingham Malaysia.
In order to verify the effects of integrating supercapacitor into the system, the
vehicle was tested with and without the ultracap module as shown in the figure
below. It should be noted that the schematic below is simplified for better
understanding. In reality the switch is a high current contactor which is controlled
by a 12V relay coil, also protection diodes, fuses have been left out of the
schematic.
Fig 60: Testing and Data Collection With and Without Supercapacitor
121
Timestamp Alt (m) GPS State Spd (kph) kph (mps) Trip (km) V_scap (V) V_bat (V) Iscap (A) Ibat (A)
17:18.0 38.4 1 0 0 0.0032 49 50.6 0 2.6
17:21.0 39.9 1 0 0 0.0032 48.9 50.5 0 6.9
17:24.0 44.9 1 0 0 0.0032 49 49 0 32.5
17:27.0 46.7 1 2.3 0.6 0.0041 48.9 46.5 0 84.1
17:30.0 45.4 1 8.9 2.5 0.0088 48.9 49.7 0 10.1
17:33.0 46.5 1 16.2 4.5 0.0192 49 49 0 26
17:36.0 45 1 17.4 4.8 0.0331 48.9 50.5 0 2.5
17:39.0 43.3 1 17.4 4.8 0.0477 48.9 50.6 0 1.6
17:42.0 45.3 1 12.4 3.5 0.0601 48.9 50.5 0 1.6
17:45.0 43.8 1 11.5 3.2 0.0701 48.9 50.6 0 2.4
17:48.0 43.6 1 11.9 3.3 0.0798 48.9 48.2 0 54
17:51.0 45.5 1 7.4 2 0.0878 48.9 49 0 15
17:54.0 46.9 1 14.4 4 0.099 48.9 48.7 0 34.7
17:57.0 48.1 1 18.6 5.2 0.1096 48.9 47.7 0 45.1
18:00.0 49 1 20.8 5.8 0.1261 48.9 50.4 0 2.5
18:03.0 48.2 1 23.3 6.5 0.1445 48.9 45.7 0 88.2
18:06.0 48.6 1 19.6 5.5 0.1624 48.9 48.8 0 33.1
18:09.0 50.2 1 23 6.4 0.1801 48.9 50.2 0 2.2
18:12.0 53.1 1 23.9 6.6 0.1995 48.9 48.7 0 33.6
18:15.0 53.8 1 25 6.9 0.22 48.9 48.3 0 45
18:18.0 53.6 1 24.3 6.7 0.2405 48.8 47.8 0 51.8
18:21.0 52.5 1 22 6.1 0.2598 48.8 47.4 0 58.4
The prototype electric kancil was driven through a specific route (figure 59) with
the switch in the OFF position i.e. battery alone, to acquire reference data for
comparison. This was repeated, but with the switch ON. Although the test driver
simulated the same driving pattern with or without the supercapacitor, the status
of traffic congestion and receiving stop light signals slightly varied through the test
time.
Fig 61: Snapshot of the data obtained from the logging system at 0.33Hz
Figure 61 above represents the time stamped raw drive data which is obtained
from the data logging device via the USB storage. The accumulated trip in km was
calculated by integrating the GPS speed over the time step between each sample.
This was achieved by some custom programming code. A MATLAB ® program was
written to calculate certain parameters from the above real world drive cycle in
order to characterize it and compare it with existing standard drive cycles.
122
CHAPTER 4: RESULTS & DISCUSSION
This chapter begins by presenting and describing the preliminary results of the
electric vehicle conversion powered by a deep cycle lead acid battery and a
supercapacitor module.
BATTERY/SUPERCAP IMPLEMENTATION
(RESULTS)
The UNMC drive cycle characterisitcs
Battery ONLY
Battery + Supercap
Battery + Supercap in a mixed cycle
DISCUSSION (COMPARING RESULTS)
Peak currents
Battery Voltage drop
Acceleration
Range test
DRIVE CYCLE ANALYSIS
Dividing the UNMC drive cycle into
microtrips
Statistical features of each microtrip
Predicting load demand?
123
4.1 BATTERY-SUPERCAPACITOR IMPLEMENTATION
This section presents the results of the integration of supercapacitor module into a
converted city electric vehicle. Raw data from the on-board data acquisition system
was collected as shown in the previous chapter. By running custom routines on
MATLAB software, basic parameters such as altitude, drive cycle data, current,
voltages etc. are plotted as well as some derived quantities such as power in
kilowatts, acceleration and jerk which are visualised.
A aキ┝WS Sヴキ┗キミェ ヴラ┌デW ┘;ゲ ;SラヮデWS キミ デエW ┌ミキ┗Wヴゲキデ┞げゲ ヮヴWマキゲWs which will be referred
to as the UNMC cycle. A total of 5 trips shown in figures 62 to 66 below, were
made. Each trip consists of approximately 10 cycles of the UNMC route which
amounts to a total distance of 10.4 km per trip. GPS altitude data collected showed
a variation from 53m to 96m with an average of 76m above sea level. It is
important to note that altitude data from GPS is subject to error (10 に 20 m
accuracy) [104], hence it is only used to prove that the test route was consistent
with all test trips made.
Due to existing local traffic regulations, an electric converted vehicle is not road
ノWェ;ノが ゲラ デWゲデキミェ ┘;ゲ SラミW ┘キデエキミ デエW ┌ミキ┗Wヴゲキデ┞げゲ campus. This route is very similar
to speed restricted driving conditions such as in residential areas with a lot of speed
bumps. Trips were done at different times of the day and also different days of the
week to see the effect on driving pattern. However, the same driver was
maintained in all cases.
130
Table 6: Summary of the main characteristics of the UNMC drive cycle for 5 different trips
TRIP 1 TRIP 2 TRIP 3 TRIP 4 TRIP 5
1 Total time
(seconds)
2406 2181 2603 1538 1698
2 Total Distance
(km)
10.46 10.44 10.82 10.52 10.20
3 Altitude (m)
Min|Mean|Max 50|75.8|90.4 60.3|80.6|100 48.1|75.8|93.7 58|78.8|103.6 49|68.5|90.6
4 Average Speed
(kph)
15.63 17.23 12.98 17.01 17.01
5 Max. Speed
(kph)
31.3 36.4 37.6 37.8 37.8
6 Stop + idle
(%)
9.6 0.7 12.54 1.68 5.29
7 1< Speed < 10kph
(%)
15.9 16.9 26.6 18.2 15
8 10 < Speed < 20kph
(%)
34.7 45.5 38.5 45.4 38.1
9 20 < Speed < 45kph
(%)
39.7 36.9 22.4 34.7 41.6
10 Speed > 45kph
(%)
0 0 0 0 0
131
Table 6 above summarizes the main characteristics of the UNMC drive cycle. For
over 70% of the total drive cycle, speeds of between 10 kmh-1
and 45 kmh-1
were
recorded and none above. Stop plus Idling time recorded was up to 12.5% which
could vary depending on amount of traffic on campus at the time of testing. Taking
a look at figure 67 above, trips 1 to 5 show very similar frequency distribution
histograms except for the idling time i.e. between 0 and 5kph. This is particularly
evident in trips 1 and 3 where the idling time is 9.6% and 12.5% respectively.
Maximum speed recorded was about 38 km/h in all trips. In the real world, this
could represent a driving trip in a residential community from say the house to the
grocery shop encountering a lot of speed bumps on the way.
132
4.1.1 Results: Battery Only (Supercapacitor OFF)
The electric kancil was driven over a fixed route as described in the section above,
solely powered by the 48V, 225Ah (20-hour rating) Trojan T105 deep-cycle lead acid
battery pack. The results of the power flow from the battery pack to the motor and
vice versa is summarised in the table below. It should be noted that the battery
pack is charged to 100% state of charge before carrying out the test.
Initial Battery Voltage 49.6 V
Average Battery Voltage Drop 2.7 V
Max. Voltage Drop 9.9 V (47.4 37.6)
Average Current 43.8 A
Peak Current 228.1 A
Average Instantaneous Power 1934.7 W
Peak Instantaneous Power 8553.8 W
Average Acceleration 0.37 ms-2
Max. Acceleration 1.37 ms-2
Average deceleration 0.38 ms-2
Max. deceleration 2.01 ms-2
Table 7: Summary of results (battery only)
Figures 67 to 74 are graphical plots and frequency distribution charts of battery
voltage, battery current (drawn by the load while driving), the instantaneous power
and the acceleration of the vehicle. The supercapacitor module was disconnected
via a high power contactor which operates on a 12V dc coil powered by the vehicles
accessory circuit. A maximum voltage drop of 9.9V was recorded. At that point, it
133
resulted in a dip from 47.4V to a record low of 37.6V which drew a current of 228.1
A from the batter┞ ヮ;Iニく AIIラヴSキミェ デラ デエW Tヴラテ;ミ Tヱヰヵ マ;ミ┌a;Iデ┌ヴWヴげゲ S;デ;ゲエWWデが
the lower voltage limit of each cell in the battery should not be below 1.75V at all
times during constant discharge [114]. Per battery, this amounts to 5.25V and the
WミデキヴW ヮ;Iニげゲ ノラ┘Wヴ ┗ラノデ;ェe limit is 42V. Therefore, any voltage below this limit is
considered to be a severe deep discharge which is detrimental to the health of the
battery in the long run. More so, these deep discharges are sudden and haphazard
depending on the acceleration demand of the vehicle and last only for a couple (3-
5) of seconds. This adds more strain on the battery.
An average current of 43.8 A was recorded for this particular trip which generally
produced an average voltage drop of 2.7 V during the trip. An average acceleration
of 0.37ms-2
up to a maximum of 1.37 ms-2
was recorded. The New York drive cycle
(NYCC) which features a mainly low speed, idle and stop-go condition has an
average acceleration of 0.62ms-2
and a maximum of 2.68 ms-2
. Due to the low
voltage used in the experimental vehicle, the acceleration is limited. The aim of the
first phase of this research work was to investigate the effect of integrating a
supercapacitor system into an electric vehicle to act as a power buffer system. The
average instantaneous power delivered by the battery was approximately 1.9KW
while it peaked to a maximum of 8.6KW.
135
Fig 69: Frequency Distribution of Battery Voltage
Bin Average (V) 38.2 39.5 40.8 42.1 43.4 44.7 46.0 47.3 48.6 49.9
Freq (%) 0.3 1.4 3.2 4.3 7.1 9.3 9.9 12.1 26.1 26.4
Table 8: Frequency table of Battery Voltage
137
Fig 71: Frequency Distribution of Battery Current
Bin Average(A) 0.6 24.5 48.5 72.4 96.4 120.3 144.3 168.2 192.2 216.1
Freq (%) 42.4 13.7 12.2 10.0 8.8 6.3 3.6 1.4 1.1 0.4
Table 9: Frequency table of Battery Current
139
Fig 73: Frequency Distribution of Instantaneous Power
Bin Average (KW) -0.1 0.8 1.7 2.6 3.5 4.4 5.4 6.3 7.2 8.1
Freq (%) 38.6 12.4 9.7 10.4 8.9 8.2 6.6 2.9 1.0 1.2
Table 10: Frequency table of Instantaneous Power delivered by Battery
141
Fig 75: Frequency Distribution of Vehicle Acceleration
Bin Average (ms-2
) -1.8 -1.5 -1.2 -0.8 -0.5 -0.2 0.2 0.5 0.9 1.2
Freq (%) 0.1 0.5 3.2 4.7 13.5 28.2 27.7 15.2 6.0 0.8
Table 11: Frequency table of Vehicle Acceleration
142
4.1.2 Results: Battery + Supercapacitor in Parallel
(48.3 45.8)
Battery Supercap
Initial Voltage (V) 49.7 49.7
Average Voltage Drop (V) 0.51 0.48
Max. Voltage Drop during Drive Cycle (V) 2.5 2.2
Average Current delivered by : (A) 33.10 0.44
Peak Current delivered by : (A) 111.2 131.2
Average Instantaneous Power delivered by: (W) 1530.8 16.8
Average Instantaneous Power delivered by Bat +
Supercap (W)
1547.6
Peak Instantaneous Power delivered by: (W) 4714.9 5956.9
Peak Instantaneous Power delivered by Bat +
Supercap : (W)
9356.3
Average Acceleration (ms-2
) 0.44 -
Max. Acceleration (ms-2
) 2.57 -
Average deceleration (ms-2
) 0.49 -
Max. deceleration (ms-2
) 2.78 -
Table 12: Summary of results (battery + supercapacitor)
Figures 71 to 75 are graphical plots of battery voltage, battery current (drawn by
the load while driving), the instantaneous power and the acceleration of the
vehicle. The supercapacitor module was connected in parallel to the battery pack
via a high power contactor which operates on a 12V dc coil powered by the
143
┗WエキIノWげゲ accessory circuit. A switch attached to the dashboard of the vehicle
triggers this contactor. As suggested by Pay et al [31], the supercapacitor was pre
Iエ;ヴェWS デラ デエW H;デデWヴ┞げゲ デWヴマキミ;ノ ┗ラノデ;ェW ふヴΒVぶ HWfore use. This is to ensure the
aラヴマWヴげゲ ┗ラノデ;ェW キゲ デキWS デラ デエW ノ;デデWヴげゲ ┗ラノデ;ェW ;デ ;ノノ デキマWゲく A マ;┝キマ┌マ H;デデWヴ┞
voltage drop of 2.5V was recorded. At that point, it resulted in a dip from 48.3V to
45.8V which drew a current of 111.2 A from the battery pack. This is a 49%
reduction in peak current drawn as compared to battery alone. An average current
of 33.1 A was recorded for this particular trip which generally produced an average
voltage drop of only 0.51 V during the trip. An Increase in acceleration was
recorded with an average acceleration of 0.44ms-2
(18.9% increase) up to a
maximum of 2.78 ms-2
(102.9% increase).
Shifting our attention to the supercapacitor module, we compare the figures 68 and
72 showing the current plots. In figure 68, large/sudden spikes or surges are
observed to be drawn from the battery pack, 40% of the current drawn from the
battery was above the 50A average while 17% of that current was above 100A. In
figure 72 (blue plot indicates battery current), only 21% of the battery current
logged was above 50A mark while a remarkable 0.38% of the 21% was above the
100A mark. This means that for 79% of the trip the battery provided less than 50A
which is not detrimental to its health. In the long run, this will result in prolonged
battery life. On the other hand, looking at the red plot (supercapacitor current), the
entire peak current demands (spikes) come from the buffer supercapacitor module;
up to a peak of 131.2 A. However there is a significant amount of negative current,
which indicates that its interaction (charging) with the battery pack is key to the
144
WミデキヴW エ┞HヴキS ゲ┞ゲデWマげゲ ;Hキノキデ┞ デラ マWWデ デエW ノラ;S ヴWケ┌キヴWマWミデゲ ラa デエW ┗WエキIノWく TエW
average positive current drawn from the supercapacitor was found to be less than
1A throughout the whole trip.
Figure 74 plots the instantaneous total power of the hybrid power source i.e.
battery + supercapacitor in a direct parallel combination. The total power delivered
is equal to the sum of the powers delivered by the individual sources. The
combined average power was equal to 1.5KW while the peak power was 9.4KW.
This is equivalent to, or even surpasses the average power delivered by the battery
alone with the extra advantage of diverting those high current surges from the
battery pack.
146
Fig 77: Frequency Distribution of Battery + Supercapacitor Voltage
Bin Average; Battery (V) 42.6 43.4 44.2 45.0 45.8 46.5 47.3 48.1 48.9 49.7
Freq (%) 0.6 1.2 3.5 6.4 9.8 16.3 22.9 18.8 10.5 10.0
Table 13: Frequency table of Battery Voltage
Bin Average ; Supercap(V) 42.7 43.5 44.3 45.0 45.9 46.6 47.4 48.2 49.0 49.8
Freq (%) 0.5 1.3 3.3 6.5 10.0 16.8 22.6 18.6 10.3 10.0
Table 14: Frequency table of Supercapacitor Voltage
148
Fig. 79: Frequency Distribution of Battery + Supercapacitor Current
Bin Average; Battery(A) 7.9 18.8 29.8 40.5 51.4 62.3 73.2 84.0 94.9 105.8
Freq (%) 19.3 20.1 23.9 12.3 8.7 6.3 5.22 2.8 1.0 0.4
Table 15: Frequency table of Battery Current
Bin Average; Supercap(A) -70.6 -49.3 -28.1 -6.9 14.4 35.6 56.89 78.1 99.3 120.6
Freq (%) 1.3 5.8 20.6 37.8 18.1 7.4 4.3 2.1 1.8 0.8
Table 16: Frequency table of Supercapacitor Current
150
Fig 81: Frequency Distribution of Battery + Supercapacitor Instantaneous Power
Bin Average; Battery (KW) 0.4 0.8 1.3 1.7 2.2 2.6 3.1 3.6 4.0 4.5
Freq (%) 17.5 15.6 22.9 15.5 9.3 8.6 4.3 4.6 1.13 0.6
Table 17: Frequency table of Power delivered by Battery
Bin Average; Supercapacitor (KW) -3.1 -2.1 -1.1 -0.2 0.7 1.7 2.6 3.6 4.5 5.5
Freq (%) 1.6 7.8 21.8 35.6 17.1 7.0 4.3 2.0 1.9 0.8
Table 18: Frequency table of Power delivered by Supercapacitor
152
Fig 83: Frequency Distribution of Effective Instantaneous Power
Bin Average (KW) -0.3 0.7 1.7 2.7 3.8 4.8 5.8 6.8 7.8 8.8
Freq (%) 42.6 16.4 13.1 9.8 6.2 5.7 2.7 1.7 1.1 0.8
Table 19: Frequency table of Effective Power delivered by Hybrid Source
154
Fig 85: Frequency Distribution of Vehicle Acceleration
Bin Average (ms-2
) -2.5 -2.0 -1.4 -0.9 -0.4 0.2 0.7 1.2 1.8 2.3
Freq (%) 0.3 0.9 2.1 6.2 16.0 60 11.5 2.4 0.7 0.2
Table 20: Frequency table of Vehicle Acceleration
155
4.1.3 Results: Battery + Supercapacitor in Parallel (Mixed cycle)
The third part of the on-road vehicle testing experiment involves a mixed cycle i.e.
switching on the supercapacitor module during the trip. As with previous tests,
both sources are pre-charged to the same level before testing. Table 21 below
summarizes the first 400 seconds of the trip which only utilizes the battery pack to
power the electric vehicle. An average voltage drop of 1.98V was recorded while a
whopping maximum of 8V drop was recorded which lead to a dip from 47.5V to
39.5V. This is far below thW H;デデWヴ┞げゲ ゲ;aWデ┞ ミWデ ラa ヴヲV キくWく ヱくΑヵV ヮWヴ IWノノく Aミ
average current of 39.5A up to a maximum peak of 207.2A was drawn from the
battery pack.
0 400 Seconds: Battery ON, Supercap OFF
Battery
Initial Voltage (V) 49.0
Average Voltage Drop (V) 1.98
Max. Voltage Drop during Drive Cycle (V) 8.0
Average Current delivered by : (A) 39.50
Peak Current delivered by : (A) 207.2
Average Instantaneous Power delivered by: (W) 1816.2
Peak Instantaneous Power delivered by: (W) 8495.2
Average Acceleration (ms-2
) 0.29
Max. Acceleration (ms-2
) 0.92
Average deceleration (ms-2
) 0.49
156
Max. deceleration (ms-2
) 1.42
Table 21: Summary of results (Mixed cycle I)
While driving the electric vehicle, the supercapacitor module was switched on at
401 seconds till the end of the trip (see figure 86 below). Table 22 summarises the
results which have been plotted in figures 86 to 90 below. An average voltage drop
of 0.87V for the battery and 0.78V for the supercapacitor was recorded while a
maximum of approximately 3V drop was recorded for both. This keeps the battery
within the safety zone (> 42V) unlike during the first 400 seconds of the drive cycle.
Figure 86 shows the voltage plots for both sources. The dashed green line divides
the plot into two parts; battery ON, Supercap OFF and Battery ON Supercap ON.
The difference between these two situations is very evident as the battery voltage
is haphazard for the first 400 seconds and later is held to a more stable discharge
rate by the supercapacitor module. The average current drawn from the battery
pack remains the same as expected (39.3 A) while the peak demand drastically
drops (207.2A to 89.9 A). On the other hand, the supercapacitor takes full
responsibility for fulfilling peak power demands. 195A peak was recorded while
only an average current of 0.55A was drawn from the supercapacitor. By comparing
tables 21 and 22 respectively in terms of power delivered to the load, single source
(battery only) was able to provide an average power of 1.8KW and a peak of 8.5KW
while the hybrid source (battery + supercap) was able to deliver average power on
par (1.8KW) and even surpass the peak power demand (12.5KW). An Increase in
acceleration was recorded with an average acceleration of 0.29ms-2
to 0.87ms-2
and
a maximum of 0.92 ms-2
to 1.25 ms-2
(see figure 90).
157
401 2406 Seconds: Battery ON, Supercap ON
Battery Supercap
Initial Voltage (V) 48.7 48.9
Average Voltage Drop (V) 0.87 0.78
Max. Voltage Drop during Drive Cycle (V) 3.2 2.9
Average Current delivered by : (A) 39.30 0.55
Peak Current delivered by : (A) 89.9 195.8
Average Instantaneous Power delivered by: (W) 1829.0 33.8
Peak Instantaneous Power delivered by: (W) 4049.9 9300.5
Average Instantaneous Power delivered by Bat +
Supercap: (W)
1795.2
Peak Instantaneous Power delivered by Bat +
Supercap : (W)
12520
Average Acceleration (ms-2
) 0.87 -
Max. Acceleration (ms-2
) 1.25 -
Average deceleration (ms-2
) 0.49 -
Max. deceleration (ms-2
) 1.48 -
Table 22: Summary of results (Mixed Cycle II)
163
4.1.4 Results: Range test
Range anxiety is a major stumbling block to the popularity of EVs. There is a
constant concern of running out of energy and being stranded on the way. This is
caused by the limited amount of cruising range of the energy source. For the
purpose of this research work, we are constrained by the type of battery pack used
i.e. Trojan T105 deep cycle lead acid batteries. A range test was carried out to
SWデWヴマキミW デエW WaaWIデ ラa デエW ゲ┌ヮWヴI;ヮ;Iキデラヴ マラS┌ノW ラミ デエW ┗WエキIノWげゲ ヴ;ミェW ヮWヴ
charge. The same route was chosen i.e. the UNMC drive cycle and the same driver
except for varying traffic conditions which generally could not be helped at the time
of this experiment. In all cases the battery is charged to 100% SOC and the total
range is estimated after the battery SOC swings to 50%. Table 23 below summarizes
the results obtained from the range test.
TOTAL RANGE PER CHARGE (KM)
TEST NO BATTERY ONLY BATTERY+ SUPERCAPACITOR % IMPROVEMENT
1 35.7 44.1 25.5
2 33.1 40.3 21.8
3 32.4 39.9 23.1
4 33.2 40.9 23.2
5 29.5 38.2 29.5
6 30.2 39.4 30.5
7 29.0 38.1 31.3
8 33.4 39.6 18.6
9 29.3 38.9 32.8
10 29.8 38.3 28.5
Table 23: Summary of the Range Test for electric kancil
From the results above, the battery can power the electric vehicle solely from
between 29km to 35.7km with an average range per charge of 28.2km. However,
with the supercapacitor module directly in parallel with the battery pack, there is a
significant increase in range per charge (between 21.8% and 32.8%), which resulted
164
in an average increase in range per charge to 39.8km. It should be noted that both
battery and supercapacitor are charged to 100% SOC before testing.
165
4.2 UNMC DRIVE CYCLE ANALYSES
This section is concerned with the analysis of the UNMC drive cycle for the electric
vehicle conversion. A drive cycle as we know it, is merely a representation of how
any vehicle is driven (i.e. speed versus time). Actual On-the-road drive cycles are
important for battery energy consumption and sizing studies in electric vehicles and
they represent driving patterns much better than synthesized drive cycles [114].
Even though they are difficult to implement on standard dynamometer chassis and
are plagued by poor repeatability issues, they are able to provide the most accurate
situation in terms of power flow between battery pack, supercapacitor and the
vehicles load demand for this research work. With the on-board data acquisition
system setup for the electric vehicle, we are able to see firsthand, the interaction
between the battery and the supercapacitor module during a drive cycle.
We will divide the UNMC drive cycle as shown in the previous section into
microtrips and analyse each trip based on some statistical features which have been
determined by researchers as described in the literature review of this thesis. [116]
describes a subsection of a drive cycle consisting of all data points between two idle
periods as a driving pulse which we refer to as microtrip. They further go on to
describe it as consisting of an acceleration phase, cruising phase and a subsequent
phase deceleration phase. However, the time window size for each microtrip will
vary based on the driving scenario. For example, idle + stop and go scenarios will
have a small time window while highway scenarios will have a larger window size. A
fixed time window may result in misrepresentation of actual driving scenario. A
small window size may not be able to capture the entire acceleration and
166
deceleration event, while a large window size may record multiple events. The
figure below shows UNMC drive cycle divided into 52 microtrips; 11 of which are
shown.
Fig 91: Dividing the UNMC Drive Cycle into Microtrips
For each microtrip, 14 statistical features were calculated by a custom MATLAB
program (see appendix). The results are shown in tables 24 to 28 below. Apart from
the features described in [61]-[66] of the literature review, two additional features,
the maximum positive jerk and the average positive jerk are calculated. This is to
observe the driver aggressiveness which has an effect on amount of current drawn
[64]. Columns highlighted in red represent a stop condition while columns in yellow
represent stop and go i.e. < 10kph.
167
Table 24: Statistical features of UNMC drive cycle: Microtrips 1:10
Microtrip 1 Microtrip 2 Microtrip 3 Microtrip 4 Microtrip 5 Microtrip 6 Microtrip 7 Microtrip 8 Microtrip 9 Microtrip 10
Trip Time (s) 14 10 38 44 20 44 18 10 26 28
Max. Speed (kph) 16.1 16.7 31.9 9.3 26.5 27.6 24.8 6.9 16.1 15.7
Max. Acceleration (m/s2) 0.78 1.39 1.79 0.90 1.36 1.82 1.58 0.04 0.93 0.49
Max. Decceleration (m/s2) -1.31 -1.31 -1.90 -1.29 -1.58 -1.88 -1.17 -1.11 -0.79 -0.65
Max. +ve Jerk (m/s3) 0.20 1.35 0.90 1.10 0.92 1.60 0.96 0.58 0.47 0.38
Mean Speed (kph) 6.93 10.44 18.99 3.23 14.83 17.57 17.37 6.84 10.38 10.15
Mean Acceleration (m/s2) 0.45 1.39 0.60 0.28 1.10 0.68 0.90 0.04 0.33 0.26
Std. Dev. Acceleration 0.21 0.00 0.58 0.27 0.29 0.64 0.66 0.00 0.43 0.15
Mean Decceleration (m/s2) -1.31 -0.63 -0.69 -0.25 -0.80 -0.54 -0.50 -1.11 -0.39 -0.28
Mean +ve Jerk (m/s3) 0.50 0.31 0.33 0.17 0.34 0.44 0.58 0.04 0.18 0.21
% idle (stops) 14.29 0.00 0.00 4.55 0.00 0.00 0.00 0.00 0.00 0.00
% < 10kph 57.14 60.00 21.05 95.45 30.00 18.18 22.22 100.00 61.54 42.86
% 10 < V< 20kph 28.57 40.00 36.84 0.00 40.00 22.73 33.33 0.00 38.46 57.14
% 20< V< 45kph 0.00 0.00 42.11 0.00 30.00 59.09 44.44 0.00 0.00 0.00
168
Table 25: Statistical features of UNMC drive cycle: Microtrips 11:20
Microtrip 11 Microtrip 12 Microtrip 13 Microtrip 14 Microtrip 15 Microtrip 16 Microtrip 17 Microtrip 18 Microtrip 19 Microtrip 20
Trip Time (s) 26 34 28 62 30 22 18 20 286 28
Max. Speed (kph) 16.80 32.60 21.30 36.90 25.40 16.40 9.50 17.00 7.60 18.40
Max. Acceleration (m/s2) 0.81 2.57 0.75 1.13 1.25 0.46 0.35 0.85 0.51 0.50
Max. Decceleration (m/s2) -0.74 -2.24 -0.94 -2.28 -2.28 -0.57 -0.39 -1.06 -0.51 -0.74
Max. +ve Jerk (m/s3) 0.37 1.28 0.63 0.57 1.28 0.24 0.20 0.42 0.26 0.35
Mean Speed (kph) 12.47 13.08 13.95 20.54 14.65 11.41 7.36 8.48 0.54 10.93
Mean Acceleration (m/s2) 0.27 1.14 0.56 0.41 0.76 0.35 0.18 0.55 0.17 0.24
Std. Dev. Acceleration 0.27 1.29 0.15 0.32 0.45 0.16 0.24 0.41 0.15 0.16
Mean Decceleration (m/s2) -0.33 -0.79 -0.43 -0.94 -0.67 -0.22 -0.14 -0.41 -0.19 -0.52
Mean +ve Jerk (m/s3) 0.19 0.52 0.43 0.38 0.43 0.09 0.04 0.03 0.09 0.18
% idle (stops) 0.00 0.00 0.00 0.00 0.00 0.00 0.00 10.00 80.42 0.00
% < 10kph 23.08 29.41 28.57 9.68 46.67 36.36 100.00 60.00 19.58 42.86
% 10 < V< 20kph 76.92 64.71 50.00 35.48 26.67 63.64 0.00 30.00 0.00 57.14
% 20< V< 45kph 0.00 5.88 21.43 54.84 26.67 0.00 0.00 0.00 0.00 0.00
169
Table 26: Statistical features of UNMC drive cycle: Microtrips 21:30
Microtrip 21 Microtrip 22 Microtrip 23 Microtrip 24 Microtrip 25 Microtrip 26 Microtrip 27 Microtrip 28 Microtrip 29 Microtrip 30
Trip Time (s) 110 12 132 26 24 36 56 16 24 120
Max. Speed (kph) 20.70 14.60 23.40 19.90 27.10 35.60 26.20 21.40 23.10 26.80
Max. Acceleration (m/s2) 0.67 0.88 1.06 2.03 1.54 1.35 1.28 1.21 0.94 0.93
Max. Decceleration (m/s2) -1.00 -0.58 -1.33 -1.75 -2.06 -2.25 -1.85 -1.07 -0.92 -1.21
Max. +ve Jerk (m/s3) 0.50 0.34 0.90 1.26 0.77 1.13 1.13 0.60 0.47 0.86
Mean Speed (kph) 9.60 10.25 14.93 9.65 18.33 20.70 14.22 14.95 14.38 17.74
Mean Acceleration (m/s2) 0.24 0.30 0.47 0.84 0.65 0.54 0.59 0.99 0.77 0.43
Std. Dev. Acceleration 0.23 0.39 0.33 0.85 0.63 0.44 0.51 0.31 0.26 0.27
Mean Decceleration (m/s2) -0.27 -0.56 -0.51 -0.77 -0.70 -0.93 -0.64 -0.52 -0.44 -0.45
Mean +ve Jerk (m/s3) 0.17 0.39 0.24 0.67 0.32 0.38 0.36 0.37 0.13 0.31
% idle (stops) 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
% < 10kph 70.91 50.00 16.67 84.62 25.00 16.67 39.29 37.50 41.67 13.33
% 10 < V< 20kph 27.27 50.00 63.64 15.38 16.67 16.67 32.14 25.00 25.00 46.67
% 20< V< 45kph 1.82 0.00 19.70 0.00 58.33 66.67 28.57 37.50 33.33 40.00
170
Table 27: Statistical features of UNMC drive cycle: Microtrips 31:40
Microtrip 31 Microtrip 32 Microtrip 33 Microtrip 34 Microtrip 35 Microtrip 36 Microtrip 37 Microtrip 38 Microtrip 39 Microtrip 40
Trip Time (s) 54 52 18 22 76 48 60 48 62 28
Max. Speed (kph) 23.80 26.40 21.40 17.40 26.30 21.20 23.60 21.00 10.30 17.80
Max. Acceleration (m/s2) 1.92 1.61 0.75 0.60 0.68 1.08 1.46 0.81 0.44 1.03
Max. Decceleration (m/s2) -1.28 -1.76 -1.57 -0.85 -1.74 -1.33 -1.65 -1.26 -0.51 -0.85
Max. +ve Jerk (m/s3) 0.96 0.88 0.50 0.31 0.68 0.65 0.83 0.63 0.63 0.65
Mean Speed (kph) 17.18 15.57 13.67 12.61 18.42 14.93 16.34 13.23 6.04 13.29
Mean Acceleration (m/s2) 0.57 0.55 0.65 0.42 0.35 0.54 0.32 0.33 0.26 0.43
Std. Dev. Acceleration 0.60 0.56 0.15 0.21 0.25 0.43 0.36 0.23 0.15 0.37
Mean Decceleration (m/s2) -0.52 -0.50 -0.56 -0.34 -0.61 -0.46 -0.56 -0.32 -0.20 -0.51
Mean +ve Jerk (m/s3) 0.33 0.43 0.03 0.22 0.24 0.29 0.21 0.24 0.02 0.34
% idle (stops) 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 12.90 0.00
% < 10kph 22.22 34.62 44.44 27.27 13.16 16.67 13.33 16.67 83.87 21.43
% 10 < V< 20kph 29.63 26.92 22.22 72.73 39.47 79.17 46.67 70.83 3.23 78.57
% 20< V< 45kph 48.15 38.46 33.33 0.00 47.37 4.17 40.00 12.50 0.00 0.00
171
Table 28: Statistical features of UNMC drive cycle: Microtrips 41:52
Microtrip 41 Microtrip 42 Microtrip 43 Microtrip 44 Microtrip 45 Microtrip 46 Microtrip 47 Microtrip 48 Microtrip 49 Microtrip 50 Microtrip 51 Microtrip 52
Trip Time (s) 86.00 132.00 42.00 56.00 38.00 44.00 16.00 114.00 38.00 58.00 214.00 60.00
Max. Speed (kph) 9.30 37.60 25.90 23.70 27.50 26.10 9.20 27.70 23.60 23.60 33.90 18.50
Max. Acceleration (m/s2) 0.31 1.11 2.47 1.21 0.96 1.38 0.36 0.96 1.50 0.94 1.75 1.29
Max. Decceleration (m/s2) -0.85 -2.78 -2.78 -2.43 -1.61 -1.10 -0.36 -1.29 -1.65 -1.47 -1.88 -1.29
Max. +ve Jerk (m/s3) 0.41 0.41 2.63 1.22 1.22 0.69 0.36 0.63 0.75 0.74 0.95 0.65
Mean Speed (kph) 2.38 16.16 14.75 15.95 17.68 14.90 7.45 16.63 17.98 13.56 18.05 9.35
Mean Acceleration (m/s2) 0.13 0.51 0.55 0.48 0.45 0.47 0.26 0.34 0.54 0.58 0.48 0.51
Std. Dev. Acceleration 0.09 0.31 0.75 0.39 0.26 0.38 0.09 0.28 0.56 0.38 0.39 0.42
Mean Decceleration (m/s2) -0.34 -0.77 -1.35 -0.73 -0.83 -0.49 -0.14 -0.43 -0.73 -0.45 -0.55 -0.65
Mean +ve Jerk (m/s3) 0.08 0.04 0.31 0.34 0.14 0.17 0.08 0.30 0.13 0.23 0.36 0.17
% idle (stops) 34.88 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 16.67
% < 10kph 65.12 13.64 38.10 14.29 21.05 31.82 100.00 21.05 21.05 41.38 10.28 33.33
% 10 < V< 20kph 0.00 59.09 38.10 75.00 31.58 45.45 0.00 45.61 26.32 41.38 54.21 50.00
% 20< V< 45kph 0.00 27.27 23.81 10.71 47.37 22.73 0.00 33.33 52.63 17.24 35.51 0.00
176
Figures 92 に 95 above represent the stem plots of the drive cycle features against
デエW マキIヴラデヴキヮゲ ラa デエW UNMC Sヴキ┗W I┞IノWく AIIラヴSキミェ デラ Jく“ Wラミげゲ Sヴキ┗キミェ デヴWミS
identifier [44], the UNMC drive cycle falls under the category of low speed cruises
whereby the average speed is under 40kph or 25mph. In order to identify the
microtrips which contain aggressive driving i.e. rapid acceleration rates that draw
peak currents from the energy source, a ratio called the acceleration feature is used
[44]. This ratio is basically the standard deviation of positive acceleration divided by
the mean positive acceleration per microtrip. This has enabled us to identify
microtrips with high acceleration rates which correspond to peak power as shown
in the step plot in figure 95 above. In order to analyse the energy contributed by
the supercapacitor module during a peak load demand, exploded view of the power
vs time graph (microtrips 2-6) is shown in the figures below. To calculate the energy
contributed by the supercapacitor during peak load demands, the area under the
power vs time plot (red shading) is found by numerical integration methods.
According to [122], the usable energy of a supercapacitor cell or module is about
75% of its total energy E. This is in line with the assumption that the voltage is
allowed to vary to half of its rated value via a suitable dc to dc converter. However,
in this project a pure parallel connection means that the terminal voltage of the
supercapacitor is always dictated by the battery packげs terminal voltage. Every
discharge phase of the supercapacitor is always followed by a subsequent charge
phase from the traction battery (negative red shaded area). Hence the
supercapacitor is utilized on the average by about 23.6% of its total capability.
177
Figure 96: Comparing the Acceleration and Jerk Plots with the Instantaneous Power by Supercap & Battery.
178
CHAPTER 5: CONCLUSION AND FURTHER WORK
This chapter concludes this thesis; the achievements are reviewed in terms of how
much of the original research objectives set out initially, was achieved. Apart from
the set objectives, this research work has contributed to the knowledge bank by
providing first hand insight into the inner workings of a parallel battery-
supercapacitor powered electric vehicle. Rather than the normal simulation studies,
a bold step was taken towards real world drive data analysis. On a positive note,
several novel ideas will emerge from the joint efforts of many small but progressive
research contributions such as this one.
TエW aキヴゲデ ラHテWIデキ┗W ┘;ゲ デラ IヴW;デW ; ゲラノキS ヮノ;デaラヴマ aラヴ デエW ┌ミキ┗Wヴゲキデ┞げゲ WノWIデヴキI ┗WエキIノW
team to carry out research and development for the present and future
researchers. A 660 cc compact city car engine was replaced with a brushless DC
motor rated at 8KW continuous and 20KW peak. The battery pack consists of eight
T105 Trojan 6V-225 Ah deep cycle lead acid battery which builds up a voltage of
48V. In addition to this, a supercapacitor module (165F, 48V) is connected in
parallel using high power contactors in order to investigate the increase in
performance criteria such as acceleration, range, battery life etc which have been
proven in various literatures via simulation studies.
The second and third objectives were to collect real world drive data from the
electric vehicle test bench cum electric car which will provide real and novel data
for an in-depth study and analysis of the interaction between a deep-cycle lead-acid
179
battery pack and a supercapacitor module in parallel powering an electric vehicle.
The findings are summarized in tables 29 and the subsequent write-ups below.
UNMC Drive Cycle Parameter Value
Total Trip Time 2085 seconds
Total Distance 10.50 km
Average Speed 16.00 kmh-1
Maximum Speed 37.80 kmh-1
% Stop/ Idling 0 に 12.5%
% 10< Speed < 20 kmh-1
34.4 に 45.5%
% 20< Speed < 45 kmh-1
22.4 に 39.7%
% Speed > 45 kmh-1
Nil
Average Acceleration 0.37 ms-2
Maximum Acceleration 1.37 ms-2
Average Deceleration 0.38 ms-2
Maximum Deceleration 2.01 ms-2
Table 29: Characteristics of the UNMC drive cycle
The UNMC drive cycle formed the basis for testing the battery にsupercapacitor
powered test vehicle. All results obtained are limited to this test cycle.
180
Improvement in Battery Life
Overall Peak currents delivered by the battery pack reduced from 228.1A to
111.2A; this is a 49% reduction.
Voltage drops experienced by the battery pack reduced from 9.9V to 2.5V.
With this, a relatively constant load profile is created for the battery pack
which is good for its life and general health.
In comparison with existing research work by Pay et al [31], results such as a
40% reduction in battery pack current and 30% improvement in dc bus
voltage regulation were obtained. However, it was observed that the state
of charge (SOC) of the supercapacitor dropped considerably after an
acceleration.
Improvement in System Average and Peak Power Capabilities
Average power demands are handled capably by the dense battery pack (1.8
KW up to 5KW) while peak power caused by acceleration of the vehicle is
diverted to the power dense supercapacitor (9.5KW up to 12.5KW).
Improvement in Acceleration
An Increase in acceleration was recorded with an average acceleration from
0.37 ms-2
to 0.44ms-2
(an 18.9% increase). The maximum acceleration
increased from 1.37 ms-2
to 2.78 ms-2
(a 102.9% increase).
Results from Shin, Donghwa, et al [33] show that an improvement in
acceleration was achieved as a result of the supercapacitors; 16.1% higher
for 0-40 km/h acceleration, 31.3% for 0-60km/h and 38.5% for 0-80km/h. In
181
terms of range, a 10.76% improvement was recorded on a fast track and a
16.67% improvement on a slow track. Fast track in this context refers to
highway speeds with longer periods of acceleration with little or no stops
(i.e. supercapacitor has little or no chance of recharging via regenerative
braking). Slow track refers to city speeds; short bursts of acceleration as well
as braking (i.e. the supercapacitor is constantly being topped off).
Improvement in Range per Charge
With the limitations of driving style and traffic condition, the increase in
average range per charge recorded was from 28.2km to 39.8km; this is a
41% increase.
ATT R&D Co. and Ness Capacitor Co. Ltd [34] reported a 15.4% increase in
range per charge.
Results of UNMC drive cycle analysis
52 microtrips were obtained from the unmc cycle and 14 parameters were
calculated including the acceleration and jerk ratios which enabled us to
identify microtrips with aggressive driving features. These microtrips were
analysed to estimate the amount of energy contributed by the
supercapacitor module towards the load demand. By numerical integration
methods, an average of 23.6% utilization was arrived at.
As a final conclusion, this research work shows that electric vehicle load profiles are
haphazard and meeting this load demand is very crucial to the vehicles
performance as perceived by the driver. As seen from figure 96, being able to
182
respond to acceleration demands will affect the performance of the vehicle. Power
electronics in terms of switching circuits are able to arbitrate power between
battery and supercapacitor but the response time is very important. Complex
control schemes, no matter how efficient or optimal they are, as long as they are
unable to fulfil this requirement will be eventually inadequate for electric vehicles.
183
FURTHER WORK
This research work formed the genesis of an entire research group dedicated to the
research and development of supercapacitors in electric vehicles. This thesis
produced novel real world experimental data that clearly shows the interaction
between a battery pack and supercapacitor module while powering a pure electric
vehicle. This hands-on approach was taken to create a strong background and
generate interest for future researchers
Given that this report contains a thorough literature survey on intelligent energy
management techniques, future work would be to capitalize on the novel data
produced in this work. Plans to develop an energy management system that
exploits the predicted power by the drive cycle analysis are in order. For example,
we can enhance the accuracy of the acceleration prediction by sensing the driver's
motion on the acceleration pedal. Besides, we may employ more accurate power
requirement models. Since these models are complex, it would be interesting to
investigate how to adapt them for real-time prediction.
184
REFERENCES
ぷヱへ Pキゲデラキ;が Gキ;ミaヴ;ミIラく さBattery operated devices and systems: From portable
electronics to industrial productsざく EノゲW┗キWヴ “IキWミIW LキマキデWSが ヲヰヰΓく
ぷヲへ Jラエミ L;ヴマキミキW Jラエミ Lラ┘ヴ┞くが さElectric Vehicle Technology Explainedざ Jラエミ WキノW┞ and Sons Ltd, 2003.
[3] Brant, Bob. Build your own electric vehicle. TAB books, 1994.
[4] Beaton, Stuart P., Gary A. Bishop, Yi Zhang, Lowell L. Ashbaugh, Douglas R.
Lawson, and Donald H. Stedman. "On-road vehicle emissions: regulations, costs,
and benefits." Science 268, no. 5213 (1995): 991-993.
ぷヵへ RラェWノキラ Aく “┌ノノキ┗;ミが さTエW TWIエミキI;ノ B;Iニェヴラ┌ミS ラa H┞HヴキS EノWIデヴキI VWエキIノWゲざが U“ Department of Energy (DoE), Office of Transportation , July 1999. URL:
HTTP://WWW.OTT.DOE.GOV/OAAT/HEV.HTML
[6] Chiasson, John, and Baskar Vairamohan. "Estimating the state of charge of a
battery." Control Systems Technology, IEEE Transactions on 13.3 (2005): 465-470.
[7] Coleman, Martin, William Gerard Hurley, and Chin Kwan Lee. "An improved
battery characterization method using a two-pulse load test." Energy Conversion,
IEEE Transactions on 23.2 (2008): 708-713.
[8] Khaligh, Alireza, and Zhihao Li. "Battery, ultracapacitor, fuel cell, and hybrid
energy storage systems for electric, hybrid electric, fuel cell, and plug-in hybrid
electric vehicles: State of the art." Vehicular Technology, IEEE Transactions on 59.6
(2010): 2806-2814
[9] Spotnitz, R. "Simulation of capacity fade in lithium-ion batteries." Journal of
Power Sources 113.1 (2003): 72-80.
[10] Buchmann, Isidor. "Will Lithium-Ion batteries power the new millennium?"
Cadex Electronics Inc., Richmond, BC, Canada (2001).
ぷヱヱへ さLW;ヴミ ;Hラ┌デ H;デデWヴキWゲくざ B;デデWヴ┞ Uミキ┗Wヴゲキデ┞が URLぎ http://batteryuniversity.com/learn/article/types of lithium ion
ぷヱヲへ Mく “く H;ノヮWヴ ;ミS Jく Cく EノノWミHラェWミ がざ“┌ヮWヴI;ヮ;Iキデラヴぎ A BヴキWa O┗Wヴ┗キW┘ざ が ヲヰヰヶ URL: http://www.mitre.org
[13] Eugenio Faggioli, Piergeorgio Rena, Veronique Danel, X Andrieu, Ronald
Mallant, Hans Kahlen, Supercapacitors for the energy management of electric
vehicles, Journal of Power Sources, Volume 84, Issue 2, December 1999, Pages 261-
269, ISSN 0378-7753, 10.1016/S0378-7753(99)00326-2.
185
[14] Ngqongwa, Lundi Vincent. Electrochemical characterization of nanostructured
SnO2 and TiO2 for potential application as dielectric materials in sulfonated-
polyaniline-based supercapacitors. Diss. Department of Chemistry, University of the
Western Cape, 2010.
[15] Miller, John M., and Richard Smith. "Ultra-capacitor assisted electric drives for
transportation." In Electric Machines and Drives Conference, 2003. IEMDC'03. IEEE
International, vol. 2, pp. 670-676. IEEE, 2003.
[16] Guidi, Giuseppe, Energy management systems on board of electric vehicles,
based on power electronics. - , 2009. - (Doctoral Theses at NTNU ; 2009:38)
[17] Adrian Schneuwly, Bobby Maher, Juergen Auer. さUltracapacitors, the New
Thinking in the Automotive Worldざく M;┝┘Wノノ TWIエミラノラェキWゲ IミIく
[18] Shah, V. A., Jivanadhar A. Joshi, Ranjan Maheshwari, and Ranjit Roy. "Review of
ultracapacitor technology and its applications." In Fifteenth National Power Systems
Conference (NPSC), IIT Bombay, p. 142. 2008.
[19] Kuperman, Alon, and Ilan Aharon. "Batteryにultracapacitor hybrids for pulsed
current loads: A review." Renewable and Sustainable Energy Reviews 15, no. 2
(2011): 981-992.
[20] Conway, B. E. Electrochemical supercapacitors: scientific fundamentals and
technological applications (POD). Kluwer Academic/plenum. New York, 1999.
[21] Young, Kwo, Caisheng Wang, and Kai Strunz. "Electric Vehicle Battery
Technologies." In Electric Vehicle Integration into Modern Power Networks, pp. 15-
56. Springer New York, 2013.
[23] Burke, Andrew, et al. "Battery Electric Vehicles: An Assessment of the
Technology and Factors Influencing Market Readiness." Public Interest Energy
Research (PIER) Program California Energy Commission (2007).
[24] Dhameja, Sandeep. Electric vehicle battery systems. Newnes, 2002.
[25] Cadar, Dorin V., Dorin M. Petreus, and Cristian A. Orian. "A method of
determining a lithium-ion battery's state of charge." Design and Technology of
Electronics Packages,(SIITME) 2009 15th International Symposium for. IEEE, 2009.
[26] Chalk, Steven G., and James F. Miller. "Key challenges and recent progress in
batteries, fuel cells, and hydrogen storage for clean energy systems." Journal of
Power Sources 159.1 (2006): 73-80.
[27] Fetcenko, M. A., et al. "Recent advances in NiMH battery technology." Journal
of Power Sources 165.2 (2007): 544-551.
186
[28] Cao, Jian. Battery/ultra-capacitor hybrid energy storage system for electric,
hybrid electric and plug-in hybrid electric vehicles. Illinois Institute of Technology,
2010.
[29] Dougal, Roger A., Shengyi Liu, and Ralph E. White. "Power and life extension of
battery-ultracapacitor hybrids." Components and Packaging Technologies, IEEE
Transactions on 25.1 (2002): 120-131.
[30] Colton, Shane. "A Simple Series Battery/Ultracapacitor Drive System for Light
Vehicles and Educational Demonstration." PヴラIく EVERげヰΓ (2009).
[31] Pay S., and Y. Baghzouz. "Effectiveness of battery-supercapacitor combination
in electric vehicles." Power Tech Conference Proceedings, 2003 IEEE Bologna. Vol. 3.
IEEE, 2003.
[32] Shin, Donghwa, et al. "Constant-current regulator-based battery-
supercapacitor hybrid architecture for high-rate pulsed load applications." Journal
of Power Sources 205 (2012): 516-524.
[33] Dixon, Juan, et al. "Electric vehicle using a combination of ultracapacitors and
ZEBRA battery." Industrial Electronics, IEEE Transactions on 57.3 (2010): 943-949.
ぷンヴへ Kキマが M;エミ “エキニが ;ミS Yラ┌ミェ Hラ Kキマく ゎATT RわDげゲ Iミ┗キデ;が ; NWキェエHラヴエララS EノWIデヴキI Vehicle with Ultracapacitors."
[35] Bentley, Paul, David A. Stone, and Nigel Schofield. "The parallel combination of
a VRLA cell and supercapacitor for use as a hybrid vehicle peak power
buffer."Journal of power sources 147.1 (2005): 288-294.
[36] Dixon, J.W.; Ortuzar, M.E.; "Ultracapacitors + DC-DC converters in regenerative
braking system," Aerospace and Electronic Systems Magazine, IEEE , vol.17, no.8,
pp. 16- 21, Aug 2002
[37] Jキミヴ┌キ Nが FWミェIエ┌ミ “が Qキミェノキ;ミ さA Study of Energy Management System of
EノWIデヴキI VWエキIノWゲざ , Vehicle Power and Propulsion Conference, 2006. VPPC '06. IEEE.
[38] Ortuzar, M.; Moreno, J.; Dixon, J.; , "Ultracapacitor-Based Auxiliary Energy
System for an Electric Vehicle: Implementation and Evaluation," Industrial
Electronics, IEEE Transactions on , vol.54, no.4, pp.2147-2156, Aug. 2007
[39] Styler, Alexander, et al. "Active management of a heterogeneous energy store
for electric vehicles." Integrated and Sustainable Transportation System (FISTS),
2011 IEEE Forum on. IEEE, 2011.
[40] Chan, C. C. "The state of the art of electric and hybrid vehicles." Proceedings of
the IEEE 90.2 (2002): 247-275.
187
[41] Chen, Zhi Hang, M. Abul Masrur, and Yi L. Murphey. "Intelligent vehicle power
management using machine learning and fuzzy logic." Fuzzy Systems, 2008. FUZZ-
IEEE 2008.(IEEE World Congress on Computational Intelligence). IEEE International
Conference on. IEEE, 2008.
[42] Murphey, Yi L., et al. "Neural learning of driving environment prediction for
vehicle power management." Neural Networks, 2008. IJCNN 2008.(IEEE World
Congress on Computational Intelligence). IEEE International Joint Conference on.
IEEE, 2008.
[43] Langari, Reza, and Jong-Seob Won. "Integrated drive cycle analysis for fuzzy
logic based energy management in hybrid vehicles." Fuzzy Systems, 2003. FUZZ'03.
The 12th IEEE International Conference on. Vol. 1. IEEE, 2003.
[44] J. S. Won, "Intelligent Energy Mangament Agent for Parallel Hybrid Electric
Vehicles," Texas A&M University, 2003, pp. 104.
[45] Ferreira, André Augusto, et al. "Energy management fuzzy logic supervisory for
electric vehicle power supplies system." Power Electronics, IEEE Transactions
on 23.1 (2008): 107-115.
[46] Yang, Jian, et al. "Forecast of driving load of hybrid electric vehicles by using
discrete cosine transform and support vector machine." Neural Networks, 2008.
IJCNN 2008.(IEEE World Congress on Computational Intelligence). IEEE International
Joint Conference on. IEEE, 2008.
[47] Huang, Xi, Ying Tan, and Xingui He. "An Intelligent Multifeature Statistical
Approach for the Discrimination of Driving Conditions of a Hybrid Electric
Vehicle." Intelligent Transportation Systems, IEEE Transactions on 12.2 (2011): 453-
465.
[48] Liaw, Bor Yann, and Matthieu Dubarry. "From driving cycle analysis to
understanding battery performance in real-life electric hybrid vehicle
operation."Journal of power sources 174.1 (2007): 76-88.
188
[49] Shi, Shuming, et al. "Study on the fuzzy clustering method of the microtrips for
passenger car driving cycle in Changchun." Vehicle Power and Propulsion
Conference, 2009. VPPC'09. IEEE. IEEE, 2009.
[50] Fu, Jianfu, Wenzhong Gao, and Liwei Song. "Principal component analysis
based on drive cycles for hybrid electric vehicle." Vehicle Power and Propulsion
Conference, 2009. VPPC'09. IEEE. IEEE, 2009.
[51] Lin, Chan-Chiao, Huei Peng, and J. W. Grizzle. "A stochastic control strategy for
hybrid electric vehicles." American Control Conference, 2004. Proceedings of the
2004. Vol. 5. IEEE, 2004.
[52] Montazeri-Gh, M., A. Ahmadi, and M. Asadi. "Driving condition recognition for
genetic-fuzzy HEV Control." Genetic and Evolving Systems, 2008. GEFS 2008. 3rd
International Workshop on. IEEE, 2008.
[53] Ji-Hui, Zhuang, Xie Hui, and Yan Ying. "Remote self-learning of driving cycle for
electric vehicle demonstrating area." Vehicle Power and Propulsion Conference,
2008. VPPC'08. IEEE. IEEE, 2008.
ぷヵヴへ LWラミ Cく Rラゲ;ヴキラ さPラ┘Wヴ ;ミS EミWヴェ┞ M;ミ;ェWマWミt of Multiple Energy Storage
ゲ┞ゲデWマゲ キミ EノWIデヴキI VWエキIノWゲざ Cヴ;ミaキWノS Uミキ┗Wヴゲキデ┞ J┌ミW ヲヰヰΑ ヮヱヰヰく
[55] Gao, Lijun, Roger A. Dougal, and Shengyi Liu. "Power enhancement of an
actively controlled battery/ultracapacitor hybrid." Power Electronics, IEEE
Transactions on 20.1 (2005): 236-243.
[56] Carter, R., and A. Cruden. "Strategies for control of a battery-supercapacitor
system in an electric vehicle." Power Electronics, Electrical Drives, Automation and
Motion, 2008. SPEEDAM 2008. International Symposium on. IEEE, 2008.
[57] Yu, Hai, Finn Tseng, and Ryan McGee. "Driving pattern identification for EV
range estimation." Electric Vehicle Conference (IEVC), 2012 IEEE International. IEEE,
2012.
[58] Zhang, Chen, et al. "Role of terrain preview in energy management of hybrid
electric vehicles." Vehicular Technology, IEEE Transactions on 59.3 (2010): 1139-
1147.
189
[59] Wang, Rui, and Srdjan M. Lukic. "Review of driving conditions prediction and
driving style recognition based control algorithms for hybrid electric
vehicles."Vehicle Power and Propulsion Conference (VPPC), 2011 IEEE. IEEE, 2011.
[60] Zhang, Chen, and A. Vahid. "Real-time optimal control of plug-in hybrid
vehicles with trip preview." American Control Conference (ACC), 2010. IEEE, 2010.
[61] Kuhler, M. "Improved driving cycle for testing automotive exhaust emissions."
(1978).
[62] Ericsson, Eva. "Independent driving pattern factors and their influence on fuel-
use and exhaust emission factors." Transportation Research Part D: Transport and
Environment 6.5 (2001): 325-345.
[63] Staackmann, Milton, Bor Yann Liaw, and David YY Yun. "Dynamic driving cycle
analyses using electric vehicle time-series data." Energy Conversion Engineering
Conference, 1997. IECEC-97., Proceedings of the 32nd Intersociety. Vol. 3. IEEE,
1997.
[64] Murphey, Yi Lu, Robert Milton, and Leonidas Kiliaris. "Driver's style
classification using jerk analysis." Computational Intelligence in Vehicles and
Vehicular Systems, 2009. CIVVS'09. IEEE Workshop on. IEEE, 2009.
[65] Carlson, T. R., and R. C. Austin. "Development of speed correction cycles."Sierra
Research, Inc., Sacramento, CA, Report SR97-04-01 (1997).
[66] Sierra Research," SCF Improvement - Cycle Development," Sierra Report No.
SR2003-06-02, 2003.
[67] Alvarez Anton, J.C.; Garcia Nieto, P.J.; Blanco Viejo, C.; Vilan Vilan, J.A.,
"Support vector machines used to estimate the battery state-of-charge," Power
Electronics, IEEE Transactions on (2013) , vol.PP, no.99, pp.1,1, 0
[68] Furey, Terrence S., et al. "Support vector machine classification and validation
of cancer tissue samples using microarray expression data." Bioinformatics16.10
(2000): 906-914.
[69] Guo, Guodong, Stan Z. Li, and Kap Luk Chan. "Support vector machines for face
recognition." Image and Vision computing 19.9 (2001): 631-638.
190
[70] Taboada, J., et al. "Creating a quality map of a slate deposit using support
vector machines." Journal of computational and applied mathematics 204.1 (2007):
84-94.
[71] Li, Xiugang, et al. "Predicting motor vehicle crashes using support vector
machine models." Accident Analysis & Prevention 40.4 (2008): 1611-1618.
[72] Hansen, Terry, and Chia-Jiu Wang. "Support vector based battery state of
charge estimator." Journal of Power Sources 141.2 (2005): 351-358.
[73] Junping, Wang, Chen Quanshi, and Cao Binggang. "Support vector machine
based battery model for electric vehicles." Energy conversion and management47.7
(2006): 858-864.
[74] Plett, Gregory L. "Extended Kalman filtering for battery management systems
of LiPB-based HEV battery packs: Part 1. Background." Journal of Power
sources 134.2 (2004): 252-261.
[75] Plett, Gregory L. "Extended Kalman filtering for battery management systems
of LiPB-based HEV battery packs: Part 2. Modeling and identification." Journal of
power sources 134.2 (2004): 262-276.
[76] Plett, Gregory L. "Extended Kalman filtering for battery management systems
of LiPB-based HEV battery packs: Part 3. State and parameter estimation."Journal of
Power sources 134.2 (2004): 277-292.
[77] Chang, Chih-Chung, and Chih-Jen Lin. "LIBSVM: a library for support vector
machines." ACM Transactions on Intelligent Systems and Technology (TIST)2.3
(2011): 27.
[78] Huang, Xi, Ying Tan, and Xingui He. "An intelligent multifeature statistical
approach for the discrimination of driving conditions of a hybrid electric
vehicle."Intelligent Transportation Systems, IEEE Transactions on 12.2 (2011): 453-
465.
ぷΑΓへ M;ヴゲエ;ノノ Bヴ;キミが さHラ┘ EノWIデヴキI C;ヴゲ Wラヴニざが Hラ┘ゲデ┌aa┘ラヴニゲが URLぎ http://auto.howstuffworks.com/electric-car2.htm
[80] Sustainability-WSが さAヴW EノWIデヴキI C;ヴゲ “┌ゲデ;キミ;HノWざが URLぎ http://www.sustainability-ed.org.uk/pages/example3-1.htm.
ぷΒヱへ B;デデWヴ┞ Uミキ┗Wヴゲキデ┞が さAヴW H┞HヴキS C;ヴゲ HWヴW デラ “デ;┞いざ URLぎ http://batteryuniversity.com/learn/article/are_hybrid_cars_here_to_stay.
191
[82] EAMEX Capacitors Japan, さ“┌ヮWヴI;ヮ;Iキデラヴざが URLぎ
http://www.eamex.co.jp/capa2_e.html
ぷΒンへ Dラミ T┌キデW がさGWデ デエW Lラ┘Sラ┘ミ ラミ Uノデヴ;I;ヮ;Iキデラヴゲざが TWIエミラノラェ┞ ヴWヮラヴデ electronic design online , November 2007. URL:
http://electronicdesign.com/Articles/ArticleID/17465/17465.html
[84] Adriaミ “IエミW┌┘ノ┞が さ DWゲキェミキミェ Pラ┘Wヴa┌ノ EノWIデヴラミキI “ラノ┌デキラミゲ ┘キデエ
Uノデヴ;I;ヮ;Iキデラヴゲざが EEデキマWゲ SWゲキェミが M;ヴIエ ヲヰヰヶく URLぎ
http://eetimes.com/design/power-management-design/4012062/Designing-
powerful-electronic-solutions-with-ultracapacitors
[85] Mく Oヴデ┌┣;ヴ さDWゲキェミが IマヮノWマWミデ;デキラミ ;ミS W┗;ノ┌;デキラミ ラa ;ミ ;┌┝キノキ;ヴ┞ WミWヴェ┞
system for electric vehicles, based on ultracapacitors and buck-boost Iラミ┗WヴデWヴざが
2005
[99] Ghosh, A., et al. "Pattern classification of fabric defects using support vector
machines." International Journal of Clothing Science and Technology 23.2/3 (2011):
142-151.
[100] Perodua Kancil Specifications URL:
http://lordq8.tripod.com/kancil_specification.htm
[101] DataTaker DT82E Technical specifications. Available online at:
http://www.datataker.com/DT82E.php
[102] CR Magnetics Split Core Hall Effect Current Transducers. Available online at:
http://www.crmagnetics.com/Products/CR-541011S--P27.aspx
[103] CR Magnetics DC VoltageTransducer. Available online at:
http://www.crmagnetics.com/Products/CR-53101120-P51.aspx
[104] 20 Channel EM-406A SiRF III Receiver with Antenna Technical Data Sheet.
Available online at:
https://www.sparkfun.com/products/465
192
[105] Gauchia, Lucia, and Javier Sanz. "Dynamic Modelling and Simulation of
Electrochemical Energy Systems for Electric Vehicles." Urban Transport and Hybrid
Vehicles (2010): 127-150.
[106] Baisden, Andrew C., and Ali Emadi. "ADVISOR-based model of a battery and
an ultra-capacitor energy source for hybrid electric vehicles." Vehicular Technology,
IEEE Transactions on 53.1 (2004): 199-205.
[107] Jinrui, N., W. Zhifu, and R. Qinglian. "Simulation and Analysis of Performance
of a Pure Electric Vehicle with a Super-capacitor." Vehicle Power and Propulsion
Conference, 2006. VPPC'06. IEEE. IEEE, 2006.
[108] Gao, David Wenzhong, Chris Mi, and Ali Emadi. "Modeling and simulation of
electric and hybrid vehicles." Proceedings of the IEEE 95.4 (2007): 729-745.
[109] Niemeier, Debbie A., Thirayoot Limanond, and Jennifer E. Morey. Data
collection for driving cycle development: evaluation of data collection protocols.
Institute of Transportation Studies, University of California at Davis, 1999.
[110] Dai, Zhen, Deb Niemeier, and Douglas Eisinger. "Driving Cycles: A New
CycleどBuilding Method that Better Represents RealどWorld Emissions." Department
of Civil and Environmental Engineering, University of California, Davis (2008).
[111] Schematic Diagram for Electric Vacuum Brake Pump System. Available online
at: http://www.zuglet.com/ev/saturn/vacuum.html
[112] Maxwell Technologies Ultracapacitor Discharge Profiles. Available online at:
http://www.maxwell.com/products/ultracapacitors/downloads
[113] Linden, David, and Thomas B. Reddy. "Handbook of batteries." New
York(2002).
[114] Trojan® T105 Deep cycle lead acid batteries, technical specifications. Available
online at:
http://www.trojanbatteryre.com/PDF/datasheets/T105_TrojanRE_Data_Sheets.pdf
ぷヱヱヵへ K;マHノWが “;ミェエヮヴキ┞; HくがTラマ Vく M;デエW┘が ;ミS GくKく “エ;ヴマ;く さ DW┗WノラヮマWミデ ラa ヴW;ノ ┘ラヴノS Sヴキ┗キミェ I┞IノWぎ C;ゲW “デ┌S┞ ラa P┌ミWが IミSキ;くざ Tヴ;ミゲヮラヴデ;デキラミ RWゲW;ヴIエ P;ヴデ Dぎ Transport and Environment 14.2 (2009): 132-140.
ぷヱヱヶへ “Iエ┘;ヴ┣Wヴが Vくが GエラヴH;ミキが Rが さDヴキ┗W C┞IノW GWミWヴ;デキラミ aラヴ DWゲキェミ Oヮデキマキ┣;デキラミ ラa EノWIデヴキI VWエキIノWゲざが Vehicular Technology, IEEE Transactions on, vol. 62,
no.1,pp.89,97, Jan 2013.
[117] M;┝┘Wノノげゲ BMODヰヱヶヵPヰヴΒ ┌ノデヴ;I;ヮ;Iキtor module, technical specifications.
Available online at:
193
http://www.maxwell.com/products/ultracapacitors/docs/datasheet_48v_series_10
09365.pdf
[118] Golden Motors International, HPM-1OkW Brushless DC Motor, Technical
specicifations. Available online at:
http://www.goldenmotor.com/eCar/HPM10KW%20(48V)%20Curve.pdf
[119] Golden Motors International, HPC500 Series -- Brushless Motor Controller,
Technical Specifications available online at:
http://www.goldenmotor.com/HPC%20Series%20Controller%20User%20Guide.pdf
[120]Holmén, Britt A., and Debbie A. Niemeier. "Characterizing the effects of driver
variability on real-world vehicle emissions." Transportation Research Part D:
Transport and Environment 3.2 (1998): 117-128.
[121]De Vlieger, I., D. De Keukeleere, and J. G. Kretzschmar. "Environmental effects
of driving behaviour and congestion related to passenger cars." Atmospheric
Environment 34.27 (2000): 4649-4655.
[122]Burke, Andrew. "Ultracapacitor technologies and application in hybrid and
electric vehicles." International Journal of Energy Research 34.2 (2010): 133-151.
197
Data Processing Routine MATLAB
load unmcdrivecycle3 figure(1); plot(unmcdrivecycle3(:,1),unmcdrivecycle3(:,7),'r'); title ('Battery + Supercapacitor current over a fixed drive cycle'); hold on plot(unmcdrivecycle3(:,1),unmcdrivecycle3(:,8),'b'); xlabel('time in seconds'); ylabel('battery current & supercap current in Amperes'); legend('supercap','battery'); hold off figure(2); plot(unmcdrivecycle3(:,1),unmcdrivecycle3(:,5),'r'); title ('Battery + Supercapacitor Voltage over a fixed drive cycle'); hold on plot(unmcdrivecycle3(:,1),unmcdrivecycle3(:,6),'b'); xlabel('time in seconds'); ylabel('battery voltage & supercap voltage in volts'); legend('supercap','battery'); hold off figure(3); plot(unmcdrivecycle3(:,1),unmcdrivecycle3(:,3),'b'); title ('drive cycle for ekancil with battery + supercap'); xlabel('time in seconds'); ylabel('GPS speed in kmh-1'); figure(4); plot(unmcdrivecycle3(:,1),unmcdrivecycle3(:,2),'b'); title ('drive cycle for ekancil with battery + supercap'); xlabel('time in seconds'); ylabel('GPS Altitude in meters(m)'); figure(5); [powerdata1]=(unmcdrivecycle3(:,5)).*unmcdrivecycle3(:,7); plot(unmcdrivecycle3(:,1),powerdata1,'r'); hold on [powerdata2]=(unmcdrivecycle3(:,6)).*unmcdrivecycle3(:,8); plot(unmcdrivecycle3(:,1),powerdata2,'b'); title ('Instantaneous Power for ekancil drive cycle'); xlabel('time in seconds'); ylabel('instantaneous power in Watts'); legend('supercap','battery'); hold off peakcurrent = max(unmcdrivecycle3(:,8)); peakpower = max(powerdata2); minVdrop= min (unmcdrivecycle3(:,6)); peakcurrentscap = max(unmcdrivecycle3(:,7)); peakpowerscap = max(powerdata1); averagepowerbat= mean(powerdata2);
198
averagepowerscap = mean(powerdata1); minVdropscap= min (unmcdrivecycle3(:,5)); maxspeed = max(unmcdrivecycle3(:,3)); [totalpower]=[powerdata2]+[powerdata1]; figure(6); plot(unmcdrivecycle3(:,1),totalpower,'g'); title ('Total Instantaneous Power (Battery + Supercap) for ekancil drive cycle'); xlabel('time in seconds'); ylabel('instantaneous power in Watts'); legend('Battery + Supercap'); %time_=num_(:,1); kph=unmcdrivecycle3(:,3); ms1=(kph*10)/36; % kph to ms1 a(1)=0; for i=1:length(ms1); %for j=1:361; V_ms(i)=ms1(i); %t(j)=time_FUDS(j); if i>1 %&& j>1; a(i)=((V_ms(i)-V_ms(i-1))/2); %end end end ms2=transpose(a); a2(1)=0; for i=1:length(ms2); %for j=1:361; V_ms2(i)=ms2(i); %t(j)=time_FUDS(j); if i>1 %&& j>1; a2(i)=((V_ms2(i)-V_ms2(i-1))/2); %end end end ms3=transpose(a2); V_bat= unmcdrivecycle3(:,6); for i=1:1315; if i>0; %&& j>1; V_drop(i)=V_bat(i)-V_bat(i+1); %end end end V_dropbat=transpose(V_drop); positiveV_dropindex= find (V_dropbat>0); %positiveV_drop= mean(V_dropbat(positiveV_dropindex)); positiveV_drop= (V_dropbat(positiveV_dropindex)); %positiveV_drop= mean(V_dropbat(positiveV_dropindex)); V_scap= unmcdrivecycle3(:,5); for i=1:1315;
199
V_dropcap(i)=V_scap(i)-V_scap(i+1); end V_dropscap=transpose(V_dropcap); positiveV_dropscapindex= find (V_dropscap>0); positiveV_dropscap= mean(V_dropscap(positiveV_dropscapindex)); %positiveV_dropscap= (V_dropscap(positiveV_dropscapindex)); %positiveV_drop= mean(V_dropbat(positiveV_dropindex)); figure(7); plot(unmcdrivecycle3(:,1),ms2,'b'); title ('drive cycle for ekancil with battery + supercap'); xlabel('time in seconds'); ylabel('acceleration in ms-2'); figure(8); plot(unmcdrivecycle3(:,1),ms3,'b'); title ('drive cycle for ekancil with battery + supercap'); xlabel('time in seconds'); ylabel('jerk in ms-3'); av_speed = mean(V_ms); max_speed = max(V_ms); std_speed = std(V_ms); max_a = max(ms2); max_decc=min(ms2); sum_acc=0; sum_decc=0; acc_count=0; decc_count=0; for i=1:length(unmcdrivecycle3) if ms2(i)>0.0; sum_acc=sum_acc+ms2(i); acc_count = acc_count + 1; end if ms2(i)<0.0 sum_decc=sum_decc+ms2(i); decc_count = decc_count + 1; end end mean_av_acc=sum_acc/acc_count; mean_av_decc=sum_decc/decc_count; idle_count=0; for i= 1: length(unmcdrivecycle3); if V_ms(i)>=0 && V_ms(i)<= 0.3; idle_count=idle_count+1; end; end; percentidle = (idle_count/length(V_ms))*100;
200
% to find the percentage of time in speed interval PLS USE "FIND" % instruction count1=0; count2=0; count3=0; count4=0; for i= 1: length(V_ms); if V_ms(i)>0.3 && V_ms(i)<= 2.7778 ; % 0 - 10km/h count1=count1+1; end; end; percent010=(count1/length(V_ms))*100; for i= 1: length(V_ms); if V_ms(i)>2.7778 && V_ms(i)<= 5.5556 ; % 10km/h - 20km/h count2=count2+1; end; end; percent1020=(count2/length(V_ms))*100; for i= 1: length(V_ms); if V_ms(i)>5.5556 && V_ms(i)<= 12.5 ; % 20km/h - 45 km/h count3=count3+1; end; end; percent2045=(count3/length(V_ms))*100; for i= 1: length(V_ms); if V_ms(i)>12.5 ; % above 45km/h count4=count4+1; end; end; percent45=(count4/length(V_ms))*100;
Drive Cycle Analysis MATLAB
clear all load trip3analysis %% microtrips for the UNMC drive cycle trip 3 % microtrip 1 utrip1=trip3analysis(4:10,2); utrip1_acc=trip3analysis(4:10,3); utrip1_jerk=trip3analysis(4:10,4); utrip1_triptime = length(utrip1)* 2; av_speed_utrip1=mean(utrip1); max_speed_utrip1=max(utrip1); max_acc_utrip1 = max(utrip1_acc); max_decc_utrip1=min(utrip1_acc); utrip1_power=trip3analysis(4:10,5); mean_power_utrip1=mean(utrip1_power); max_power_utrip1=max(utrip1_power); utrip1_acc_index = find(utrip1_acc > 0);
201
utrip1_acc_values = utrip1_acc(utrip1_acc_index); % list all the acceleration values utrip1_decc_index = find(utrip1_acc < 0); utrip1_decc_values = utrip1_acc(utrip1_decc_index); % list all the decceleration values std_acc_utrip1 = std(utrip1_acc_values); mean_acc_utrip1= sum(utrip1_acc_values)/length(utrip1_acc_values); mean_decc_utrip1=sum(utrip1_decc_values)/length(utrip1_decc_values); max_jerk_utrip1=max (utrip1_jerk); utrip1_jerk_index = find(utrip1_jerk > 0); utrip1_jerk_values = utrip1_acc(utrip1_jerk_index); % list all the postive jerk values mean_jerk_utrip1=sum(utrip1_jerk_values)/length(utrip1_jerk_values); % to find the percentage of time in speed interval PLS USE "FIND" % instruction percent010_utrip1_index = find(utrip1>0 & utrip1<=10); % 0-10 km/h percent010_utrip1_values=utrip1(percent010_utrip1_index); percent010_utrip1 = (length(percent010_utrip1_values)/length(utrip1))*100; percent1020_utrip1_index = find(utrip1>10 & utrip1<=20); % 10-20 km/h percent1020_utrip1_values=utrip1(percent1020_utrip1_index); percent1020_utrip1 = (length(percent1020_utrip1_values)/length(utrip1))*100; percent2045_utrip1_index = find(utrip1>20 & utrip1<=45); % 20-45 km/h percent2045_utrip1_values=utrip1(percent2045_utrip1_index); percent2045_utrip1 = (length(percent2045_utrip1_values)/length(utrip1))*100; percentidle_utrip1_index = find(utrip1==0); percentidle_utrip1_values = utrip1(percentidle_utrip1_index); percentidle_utrip1=(length(percentidle_utrip1_values)/(length(utrip1)))*100; feature_utrip1 = [utrip1_triptime;max_speed_utrip1;max_acc_utrip1;max_decc_utrip1;max_jerk_utrip1;av_speed_utrip1; mean_acc_utrip1;std_acc_utrip1;mean_decc_utrip1;mean_jerk_utrip1;percentidle_utrip1;percent010_utrip1;percent1020_utrip1; percent2045_utrip1]; %% microtrip 2 utrip2=trip3analysis(10:14,2); utrip2_acc=trip3analysis(10:14,3); utrip2_jerk=trip3analysis(10:14,4); utrip2_triptime = length(utrip2)* 2; av_speed_utrip2=mean(utrip2); max_speed_utrip2=max(utrip2);
202
max_acc_utrip2 = max(utrip2_acc); max_decc_utrip2=min(utrip2_acc); utrip2_power=trip3analysis(10:14,5); mean_power_utrip2=mean(utrip2_power); max_power_utrip2=max(utrip2_power); utrip2_acc_index = find(utrip2_acc > 0); utrip2_acc_values = utrip2_acc(utrip2_acc_index); % list all the acceleration values utrip2_decc_index = find(utrip2_acc < 0); utrip2_decc_values = utrip2_acc(utrip2_decc_index); % list all the decceleration values std_acc_utrip2 = std(utrip2_acc_values); mean_acc_utrip2= sum(utrip2_acc_values)/length(utrip2_acc_values); mean_decc_utrip2=sum(utrip2_decc_values)/length(utrip2_decc_values); max_jerk_utrip2=max (utrip2_jerk); utrip2_jerk_index = find(utrip2_jerk > 0); utrip2_jerk_values = utrip2_acc(utrip2_jerk_index); % list all the postive jerk values mean_jerk_utrip2=sum(utrip2_jerk_values)/length(utrip2_jerk_values); % to find the percentage of time in speed interval PLS USE "FIND" % instruction percent010_utrip2_index = find(utrip2>0 & utrip2<=10); % 0-10 km/h percent010_utrip2_values=utrip2(percent010_utrip2_index); percent010_utrip2 = (length(percent010_utrip2_values)/length(utrip2))*100; percent1020_utrip2_index = find(utrip2>10 & utrip2<=20); % 10-20 km/h percent1020_utrip2_values=utrip2(percent1020_utrip2_index); percent1020_utrip2 = (length(percent1020_utrip2_values)/length(utrip2))*100; percent2045_utrip2_index = find(utrip2>20 & utrip2<=45); % 20-45 km/h percent2045_utrip2_values=utrip2(percent2045_utrip2_index); percent2045_utrip2 = (length(percent2045_utrip2_values)/length(utrip2))*100; percentidle_utrip2_index = find(utrip2==0); percentidle_utrip2_values = utrip2(percentidle_utrip2_index); percentidle_utrip2=(length(percentidle_utrip2_values)/(length(utrip2)))*100; feature_utrip2 = [utrip2_triptime;max_speed_utrip2;max_acc_utrip2;max_decc_utrip2;max_jerk_utrip2;av_speed_utrip2;
203
mean_acc_utrip2;std_acc_utrip2;mean_decc_utrip2;mean_jerk_utrip2;percentidle_utrip2;percent010_utrip2;percent1020_utrip2; percent2045_utrip2];
.
.
.
%% microtrip52 utrip52=trip3analysis(1287:1316,2); utrip52_acc=trip3analysis(1287:1316,3); utrip52_jerk=trip3analysis(1287:1316,4); utrip52_triptime = length(utrip52)* 2; av_speed_utrip52=mean(utrip52); max_speed_utrip52=max(utrip52); max_acc_utrip52 = max(utrip52_acc); max_decc_utrip52=min(utrip52_acc); utrip52_power=trip3analysis(1287:1316,5); mean_power_utrip52=mean(utrip52_power); max_power_utrip52=max(utrip52_power); utrip52_acc_index = find(utrip52_acc > 0); utrip52_acc_values = utrip52_acc(utrip52_acc_index); % list all the acceleration values utrip52_decc_index = find(utrip52_acc < 0); utrip52_decc_values = utrip52_acc(utrip52_decc_index); % list all the decceleration values std_acc_utrip52 = std(utrip52_acc_values); mean_acc_utrip52= sum(utrip52_acc_values)/length(utrip52_acc_values); mean_decc_utrip52=sum(utrip52_decc_values)/length(utrip52_decc_values); max_jerk_utrip52=max (utrip52_jerk); utrip52_jerk_index = find(utrip52_jerk > 0); utrip52_jerk_values = utrip52_acc(utrip52_jerk_index); % list all the postive jerk values mean_jerk_utrip52=sum(utrip52_jerk_values)/length(utrip52_jerk_values); % to find the percentage of time in speed interval PLS USE "FIND" % instruction percent010_utrip52_index = find(utrip52>0 & utrip52<=10); % 0-10 km/h percent010_utrip52_values=utrip52(percent010_utrip52_index); percent010_utrip52 = (length(percent010_utrip52_values)/length(utrip52))*100;
204
percent1020_utrip52_index = find(utrip52>10 & utrip52<=20); % 10-20 km/h percent1020_utrip52_values=utrip52(percent1020_utrip52_index); percent1020_utrip52 = (length(percent1020_utrip52_values)/length(utrip52))*100; percent2045_utrip52_index = find(utrip52>20 & utrip52<=45); % 20-45 km/h percent2045_utrip52_values=utrip52(percent2045_utrip52_index); percent2045_utrip52 = (length(percent2045_utrip52_values)/length(utrip52))*100; percentidle_utrip52_index = find(utrip52==0); percentidle_utrip52_values = utrip52(percentidle_utrip52_index); percentidle_utrip52=(length(percentidle_utrip52_values)/(length(utrip52)))*100; feature_utrip52 = [utrip52_triptime;max_speed_utrip52;max_acc_utrip52;max_decc_utrip52;max_jerk_utrip52;av_speed_utrip52; mean_acc_utrip52;std_acc_utrip52;mean_decc_utrip52;mean_jerk_utrip52;percentidle_utrip52;percent010_utrip52;percent1020_utrip52; percent2045_utrip52]; %% UDDS_parameters = [feature_utrip1 feature_utrip2 feature_utrip3 feature_utrip4 feature_utrip5 feature_utrip6 feature_utrip7 feature_utrip8 feature_utrip9 feature_utrip10 feature_utrip11 feature_utrip12 feature_utrip13 feature_utrip14 feature_utrip15 feature_utrip16 feature_utrip17 feature_utrip18 feature_utrip19 feature_utrip20 feature_utrip21 feature_utrip22 feature_utrip23 feature_utrip24 feature_utrip25 feature_utrip26 feature_utrip27 feature_utrip28 feature_utrip29 feature_utrip30 feature_utrip31 feature_utrip32 feature_utrip33 feature_utrip34 feature_utrip35 feature_utrip36 feature_utrip37 feature_utrip38 feature_utrip39 feature_utrip40 feature_utrip41 feature_utrip42 feature_utrip43 feature_utrip44 feature_utrip45 feature_utrip46 feature_utrip47 feature_utrip48 feature_utrip49 feature_utrip50 feature_utrip51 feature_utrip52 ]; UDDS_parameters=transpose(UDDS_parameters); avpower_drive_cycle = [mean_power_utrip1;mean_power_utrip2;mean_power_utrip3;mean_power_utrip4;mean_power_utrip5;mean_power_utrip6;mean_power_utrip7; mean_power_utrip8;mean_power_utrip9;mean_power_utrip10;mean_power_utrip11;mean_power_utrip12;mean_power_utrip13;mean_power_utrip14;mean_power_utrip15; mean_power_utrip16;mean_power_utrip17;mean_power_utrip18;mean_power_utrip19;mean_power_utrip20;mean_power_utrip21;mean_power_utrip22;mean_power_utrip23; mean_power_utrip24;mean_power_utrip25;mean_power_utrip26;mean_power_utrip27;mean_power_utrip28;mean_power_utrip29;mean_power_utrip30;mean_power_utrip31;
205
mean_power_utrip32;mean_power_utrip33;mean_power_utrip34;mean_power_utrip35;mean_power_utrip36;mean_power_utrip37;mean_power_utrip38;mean_power_utrip39; mean_power_utrip40;mean_power_utrip41;mean_power_utrip42;mean_power_utrip43;mean_power_utrip44;mean_power_utrip45;mean_power_utrip46;mean_power_utrip47; mean_power_utrip48;mean_power_utrip49;mean_power_utrip50;mean_power_utrip51;mean_power_utrip52]; maxpower_drive_cycle = [max_power_utrip1;max_power_utrip2;max_power_utrip3;max_power_utrip4;max_power_utrip5;max_power_utrip6;max_power_utrip7; max_power_utrip8;max_power_utrip9;max_power_utrip10;max_power_utrip11;max_power_utrip12;max_power_utrip13;max_power_utrip14;max_power_utrip15; max_power_utrip16;max_power_utrip17;max_power_utrip18;max_power_utrip19;max_power_utrip20;max_power_utrip21;max_power_utrip22;max_power_utrip23; max_power_utrip24;max_power_utrip25;max_power_utrip26;max_power_utrip27;max_power_utrip28;max_power_utrip29;max_power_utrip30;max_power_utrip31; max_power_utrip32;max_power_utrip33;max_power_utrip34;max_power_utrip35;max_power_utrip36;max_power_utrip37;max_power_utrip38;max_power_utrip39; max_power_utrip40;max_power_utrip41;max_power_utrip42;max_power_utrip43;max_power_utrip44;max_power_utrip45;max_power_utrip46;max_power_utrip47; max_power_utrip48;max_power_utrip49;max_power_utrip50;max_power_utrip51;max_power_utrip52]; figure (1); stem(UDDS_parameters(:,2),'r'); title ('Comparing drive cycle features'); hold on plot(UDDS_parameters(:,6),'b'); xlabel('microtrip'); ylabel('drive cycle feature'); legend('Max. speed (m/s)','Average speed (m/s)'); hold off figure (2); stem(UDDS_parameters(:,3),'r'); title ('Comparing drive cycle features'); hold on plot(UDDS_parameters(:,7),'b'); xlabel('microtrip'); ylabel('drive cycle feature'); legend('Max. acceleration (m/s^2)','Average acceleration (m/s^2)'); hold off
206
figure (3); stem(UDDS_parameters(:,5),'r'); title ('Comparing drive cycle features'); hold on plot(UDDS_parameters(:,10),'b'); xlabel('microtrip'); ylabel('drive cycle feature'); legend('Max. jerk (m/s^3)','Average Jerk (m/s)^3'); hold off figure (4); stem(maxpower_drive_cycle,'r'); title ('Comparing drive cycle features'); hold on plot(avpower_drive_cycle,'b'); xlabel('microtrip'); ylabel('drive cycle power in W'); legend('Max Power(W)','Av. Power (W)'); hold off
207
APPENDIX C
Electric Bicycle with Supercapacitors as Peak Power Buffer
This work implements a smart boost converter to enable an electric bicycle to be powered by
a battery/supercapacitor hybrid combination. A 36V, 250W front hub motor was retrofitted
onto a normal geared bike powered by a 36V; 12Ah lithium ion phosphate battery pack. A
16.2V, 58F supercapacitor module was connected in parallel to the battery pack via a custom
made microcontroller-based boost converter which arbitrates power between the battery
and supercapacitor. The control algorithm for the boost converter was developed using a
practical approach by using various sensor inputs (battery/supercapacitor current and
voltage, bike speed) and comparing the robustness of control scheme. Also energy efficient
components were used in designing the boost converter to ensure maximum power transfer
efficiency. Based on the implemented system experimental results show an improvement in
the up-hill acceleration of the bicycle as a direct result of the boost converter being
responsive enough to harvest the extra current from the high power complementary
supercapacitor module avoiding deep discharges from the battery. This enhanced battery
life. The maximum speed remained unchanged while the improvement in range per charge
was subjective to the terrain i.e. flat land; not significant improvement, hilly terrain;
significant. However, recharging the supercapacitor via regenerative braking proved to be an
arduous task since the boost converter was not designed to be bi-directional.
I. INTRODUCTION.
For much of the world; especially places like China, India, and Sweden etc. bicycles have been a
transportation mainstay because the work place and housing areas in most of these densely
populated cities are within walking or cycling distance. This reliable yet overlooked form of
transportation has evolved over the years from simple utility bicycles to powerful geared mountain
bikes and now electric assisted bicycles or pedelecs. Environmental concerns in terms of emissions
and depleting fuel reserves has revived the electric vehicle industry and research community. China
has produced 21 million of bicycles within nine years (1997-2005). In 2005 itself, China has produced
10 million of bicycles (Alan Parker1).
Electric assisted bicycles still retain the characteristics of a conventional bicycle with an added
advantage of extra power, say when riding up a hill. This enables the elderly or not so physically fit
people to still enjoy riding a bicycle up a slope.
Batteries are the weak leak at the moment for any electrically propelled vehicle including the
bicycle. The lack of a single reasonably priced energy storage device that can simultaneously provide
high power density and high energy density has been the main stumbling block to the acceptance of
208
electric propulsion as the main form of private and public transportation (Manoj et al2, Schneuwly et
al5)Presently the only viable solution to this problem is to combine a high energy storage device such
as an electrochemical battery or fuel cell with a high power device such as an Electric Double Layer
Capacitor (EDLC) or ultracapacitor or more often called a supercapacitor (Ortuzar et al3). Usually,
some form of dc to dc converter executing an energy management control algorithm is used to
interface the battery bank and supercapacitor array to the load bus (Payman et al8). It is the aim of
this project to design a smart boost converter with a heuristic based energy management algorithm
which will optimize the power flow from the battery pack to the load.
As the name implies, a supercapacitor is a capacitor with capacitance greater than any other,
usually in excess of up to 4000 Farad. Supercapacitors do not have a traditional dielectric material
like ceramic, polymer films or aluminum oxide to separate the electrodes instead a physical barrier
made of activated carbon. A double electric field which is generated when charged, acts a dielectric.
The surface area of the activated carbon is large thus allowing for the absorption of large amount of
ions (Dougal et al4).
Advantages of Supercapacitors
I. Cell voltage determined by the circuit application not limited by cell chemistry
II. Very high cell voltages possible
III. High power density
IV. Can withstand extreme temperatures
V. Simple charging methods
VI. Very fast charge and discharge
VII. Overcharging not possible
VIII. Long life cycle
IX. Low impedance
Disadvantages/Shortcomings
I. Linear discharge voltage characteristic prevents use in some applications
II. Power only available for very short duration (short bursts of power)
III. Low capacity
IV. Low energy density
V. Voltage balancing required when banking
VI. High self discharge rate
Despite the fact that Lithium-ion batteries are the superior most power sources as compared to
other battery systems, the performance of the Li-ion battery is greatly affected when utilized under
high current discharges. Lithium ion batteries have a high energy density of about 105 J/kg; however
the power density is only around 100W/kg. As a result, the Li-ion battery cannot yield to high power
demands. The addition of super capacitors in parallel to the battery can greatly augment the
utilization of the battery especially at higher rates of discharge because of the high power density of
super capacitors (~106 W/kg). So during high power demands, the supercapacitor aids in supplying
209
the power instantly while it gets recharged by the battery and thus the run time of the battery super
capacitor hybrid is increased and higher utilization can be ensured.
Li-ion and supercapacitors are at opposite ends of the spectrum i.e. high energy density and high
power density respectively. A combination of both technologies could bring about very desirable
effects in terms of battery features. A Direct connection of the supercapacitor across the battery
terminals does reduce transient currents in and out of the battery. However, the best way to utilize
the supercapacitor bank is to be able control its energy content through a power converter (Pay &
Baghzouz5). Usually, a static bi-directional buck-boost converter is used to interface the
supercapacitor bank (connected to the boost side) with the battery pack (connected to the buck
side) because batteries work at relatively constant voltage levels while capacitors voltage is directly
related to its state-of-charge (Dixon & Ortuzar6).
A. Outline
The objective of this project is to design and build a Pulse Width Modulated (PWM) DC/DC
boost converter with the following ideal specifications:
Input Voltage: 12 - 16.2 VDC
Output Voltage: 36 - 39 VDC
Output Power: 200 に 300 W
Switching Frequency: 20 に 100 kHz
Figure 1: Overview of the proposed system
Figure 1 above shows the block diagram of the proposed system. The main power source consists
of a 36V~40V, 12Ah lithium ion Phosphate (LiFePo4) battery pack with cell balancing circuitry which
weighs 5.5kg. This can comfortably provide a continuous discharge current of 12A. The
supercapacitor consists of a 0.5kg POWERBUSRT© 16V, 58F module to be connected in parallel with
210
the battery pack via a boost converter which is designed to harvest the maximum energy from it.
The propulsion device consists of a state of the art front hub motor rated at 250W (Golden
Motors11
), eliminating the need for transmission and the losses associated with it.
Section 2 describes the sizing and design of the boost converter based on the specific energy
requirements of the previously mentioned electric assisted bicycle. Section 3 discusses the results
obtained from the implemented system using a portable data logger.
II. BOOST CONVERTER DESIGN
In order to size the boost converter appropriately, the electric bicycle was powered with the 36V
12Ah LiFePo4 batteries initially, and certain figures were collected. The track used was a mixture of
flat terrain, uphills and downhills. This is shown in the table below.
TABLE I. E-bikes performance with battery
Maximum Voltage 40.69 V
Maximum Current 18.38 A
Average Voltage 39.0 V
Average Current 8.2 A
Max. Voltage Drop 3.31 V
Maximum Power 693.0 W
Average Power 315.0 W
From the table above, the maximum voltage is 40.69V. This voltage is actually equivalent to the
total batteries voltage at full charge. Even though the datasheet mentions that each battery is only
supplied 36V, however there is some tolerance in battery which causes each battery to go up to 40V.
When the motor is running at maximum load (i.e. uphill), the maximum current drawn from the
battery is 18.38A whereas when the motor is running at constant load (usually on a flat terrain) the
average current is 8.2A. Thus, the supercapacitor is required to supply at least 10.18A to ensure that
batteries only supplies average current.
211
Figure 2: Conventional PWM Boost Converter
The conventional boost converter is shown in figure 2 above. It can be operated in two modes
which are continuous current mode (CCM) and discontinuous current mode (DCM. This converter
functions based on pulse width modulation (PWM); PWM is sent from a microcontroller/driver in
order to control the IGBT to switch on or switch off (Farzanehfard & Beyragh9).
During on time, IGBT will be turned on by setting the PWM to high. During the on time, current
from the voltage source will flow through inductor and IGBT. This results in positive voltage across
the inductor, thus inductor current is linearly increases. The capacitor will discharge through the
load resistance in order to provide continuous supply to the load.
During off time, current through the inductor cannot change instantaneously. It will try to oppose
any drop in current by reversing its electromotive force (EMF) or polarity. Thus, the inductor will be
treated as another voltage source which is in series with the input voltage. Energy stored in the
inductor will add extra voltage to the input voltage. The current will flow into the capacitor to charge
up the capacitor as well. In short, output voltage will be boosted with extra voltage from the
inductor.
A. Supercapacitor Energy Analysis
Total Energy 継脹 噺 怠態系撃態 (1) 継脹 噺 怠態 抜 のぱ 抜 なは態 噺 ば┻ねにね計蛍 岫に┻な激月岻
Assume the minimum voltage can drop to 10V, maximum voltage is 16V
Usable Energy; Eu = 岾な 伐 蝶鉄尿日韮蝶鉄尿尼猫峇 継脹 (2)
Eu = 4.524 KJ (1.26Wh)
Thus 452.4W (28.28A @ 16V) can be supplied for a period of 10seconds or 226.2W (14.14A @
16V) for 20 seconds. This is more than sufficient for an uphill acceleration keeping in mind that the
battery is able to supply an average power. However the boost converter is required to output a
voltage of between 38V~40V depending on the battery pack voltage leading to a lower output
current.
Assuming a 90% converter efficiency,
212
鶏墜通痛椎通痛 噺 ど┻ひ 鶏沈津椎通痛 (3) ど┻ひ 茅 撃沈津椎通痛 茅 沈津椎通痛 噺 撃墜通痛椎通痛 茅 墜通痛椎通痛 (4)
Ioutput = 12.89A @ Vinput =16V
Ioutput = 8.06A @ Vinput =10V
B. Component Sizing
The converter must be designed such that it is able to handle the worst case scenario. Thus, input
voltage used in the design calculation will be the minimum whereas the output voltage will be the
maximum. The design specifications are showed as below: 血 噺 などど倦茎権 (Arbitrarily chosen)
Duty cycle can be calculated by using the formula as follows: 経 噺 な 伐 蝶日韮蝶任祢禰 (5)
経 噺 な 伐 などねな 経 噺 ど┻ばのは
Since the output power is 500W and voltage is 39V, the resistive load can be calculated as 迎 噺 撃態鶏
迎 噺 ぬひ態のどど 噺 ぬ┻どねにм
Therefore, the minimum inductor value can be calculated by substitute the value as followed: 詣陳沈津 噺 帖岫怠貸帖岻鉄眺態捗
(6) 詣陳沈津 噺 ど┻ばのは 抜 岫な 伐 ど┻ばのは岻態 抜 ぬ┻どねにに 抜 などど 抜 など戴 詣陳沈津 噺 ど┻はぱねは航茎
The inductor must be selected such that it is able to handle input current. The converter is
assumed to be 90% efficient, thus the output power will be nine-tenth of the input power. 荊沈津 噺 鶏墜通痛ど┻ひ撃沈津 噺 なね┻にの畦
The selected inductor is J. Bournes and Miller 8121-RC which has 1mH inductance with maximum
current of 20A.The output capacitance reduces the output ripple voltage. The filtering capacitance is
selected by assuming the ripple voltage, ッ撃墜通痛 is equaled to 10% of the output voltage. 系陳沈津 噺 怠ッ蝶任 抜 蝶任眺 抜 経劇 (7)
系陳沈津 噺 など┻どの 抜 ぬひ 抜 ぬひぬ┻どねに 抜 ど┻ばのは 抜 ななどど 抜 など戴 系陳沈津 噺 のど航繋
213
The selected output capacitance is 100µH. Capacitors with higher capacitance will actually reduce
the ripple voltage, but higher capacitance will have a slower response time [6]. Table II below shows
the list of other components used in this design.
Figure 3: Control strategy for Boost Converter
TABLE II. Components for boost converter
Component Function 1 PIC16F877A Microcontroller Generate PWM signals based on
input sensors 2 HCPL 3120 Optocoupler Gate driver for switching device 3 IKP15N60T IGBT (20A) Switching device for converter 4 Vishay Schottky diode (30A) Prevent reverse current 5 LTS-25NP LEM current sensor Sense load current give input to
MCU
214
C. Control Algorithm
Figure 3 above describes the simple control strategy adopted for calculating the duty cycle
of the PWM pulse which is required to turn on the IGBT for the boost converter. Inputs from a Hall
Effect speed sensor and current sensors are required to turn on the converter. This ensures that the
E-bike is on and ready for an impending acceleration and also prevents unnecessary usage of the
ゲ┌ヮWヴI;ヮ;Iキデラヴげゲ ノキマキデWS WミWヴェ┞く Vラノデ;ェW デヴ;ミゲS┌IWヴ IキヴI┌キデゲ ;デ デエW キミヮ┌デ ゲキSW ふゲ┌ヮWヴI;ヮ;Iキデラヴぶ ;ミS
output side (battery/converter out) feed input signals to the microcontroller. By comparing these
two signals, the duty cycle can be appropriately adjusted. Figure 4 below shows the boost converter
along with its control circuit (a PIC16f877a microcontroller) while figure 5 shows the complete
system retrofitted onto a conventional bicycle.
Figure 4: Boost Converter with control circuit.
Figure 5: Ebike with battery supercapacitor hybrid via a boost converter.
215
Figure 6: Battery pack voltage of E-bike with only LiFePo4 as
power source
Figure 9: Battery pack voltage of E-bike with hybrid power
source
Figure 10: Current drawn from LiFePo4 battery with hybrid
power source Figure 7: Current drawn from LiFePo4 only
Figure 6: Battery pack voltage of E-bike with only LiFePo4 as
power source
Figure 8: Power Output from LiFePo4 battery
Figure 9: Battery pack voltage of E-bike with hybrid power
source
Figure 11: Power Output from LiFePo4 battery with hybrid
power source
216
III. RESULTS AND DISCUSSION.
A portable data logging device was used to collect sensor data of bike speed, battery voltage
and current, supercapacitor voltage and current. This was used to evaluate the performance of the
electric assisted bicycle in terms of maximum speed, voltage drop of battery pack, peak current
drawn from battery pack and up-hill acceleration. Figure 6 to figure 8 shows the battery parameters
when the E-bike was powered solely by the battery pack. An average voltage drop of 1.85V was
recorded while riding on flat terrain while a maximum voltage drop of 3.31V was recorded when
riding uphill. A peak current of 18.38A was recorded uphill. This is potentially dangerous for the
battery as it only able to handle a maximum of 15A continuous discharge. An Average Power of
315W and peak of 693W was subsequently recorded. Figure 9 to figure 11 plots the battery
parameters when the E-bike was powered by hybrid battery/supercapacitor power source. A
significant reduction in the maximum voltage drop from 3.31V to 1.64V occurred. This was a direct
result of the supercapacitor being able to attend to the peak power requests from the load. Peak
current from the battery pack reduced to 12.23A which was observed in figure 10. As a result, peak
power reduced to 470W. Table III below summarizes the results.
Battery Parameter
@ Peak conditions
LiFePo4 only LiFePo4 + Supercap
Volt. Drop 3.31V 1.64V
Current 18.38A 12.23A
Power 693W 470W
Max Speed from GPS
Speed sensor
28Km/h 28.8Km/h
For every instance in time, the current drawn from the battery-only power source is always
greater than the current drawn from the battery + supercapacitor power source. However, there
appears to be certain discrepancies which is caused by variation in driving pattern although data
collection for both instances where taken by following exactly the same route. The voltage profile of
the hybrid power source is smoother with less variations than the profile of the battery only power
source. Since the voltage of the supercapacitor is always tied to the battery pack (direct parallel
connection), we cannot fully unravel its true potential which is its ability to charge and discharge
very fast. A simple efficiency analysis shows that only 23% of the total energy of the supercapacitor
was used up. The remaining 77% cannot be utilized due to the direct parallel configuration. The
parallel hybrid power source has no significant effect on the maximum speed achievable over a drive
cycle except that the rider felt an improvement in uphill acceleration as compared to battery alone.
This test was just part of comprehensive tests that was scheduled to be carried out in near
future to optimize supercapacitor integration with an electric vehicle. Significant amount of
WミェキミWWヴキミェ ┘ラヴニ ゲデキノノ ヴWマ;キミゲ aラヴ デエW ラヮデキマキ┣;デキラミ WゲヮWIキ;ノノ┞ デエW さゲマ;ヴデざ Iラミデヴラノ ;ノェラヴキデエマ aラヴ デエW
TABLE III. Summary of Results with Battery + Supercapacitor
217
ゲ┌ヮWヴI;ヮ;Iキデラヴ デラ HW ゲ┘キデIエWS さラミざ ;ミS さラaaざ ;デ マラゲデ ;ヮヮヴラヮヴキ;デW Sヴキ┗キミェ ヮ;デデWヴミく Tエキゲ ┘ラ┌ノS Wミ;HノW
a wider r;ミェW ラa ┌ゲ;ェW ラa デエW ゲ┌ヮWヴI;ヮ;Iキデラヴげゲ WミWヴェ┞ ラa ┌ヮ デラ ΑヵХく
IV. CONCLUSION.
In conclusion, this work has successfully implemented a battery/supercapacitor hybrid power
source for an electric assisted bicycle using state of the art hub motor technology. A boost converter
was designed and implemented based on the energy requirements of the system.
Based on the implemented system experimental results show an improvement in the up-hill
acceleration of the bicycle as a direct result of the boost converter being responsive enough to
harvest the extra current from the high power complementary supercapacitor module avoiding deep
discharges from the battery. This enhanced battery life. The maximum speed remained unchanged.
The main battery pack was shielded from high discharge currents which would eventually enhance
its life cycle.
ACKNOWLEDGEMENTS
The author of the work would like to thank MOSTI (Ministry of Science Technology and
Innovation Grant No: 01-02-12-SF0095) Malaysia and Sahz holdings.
REFERENCES
1 Alan. A Parker, Electric Power-Assisted Bicycles Reduce Oil Dependence and Enhance the Mobility of the Elderly electric
power assisted bicycle. Presented at 29th Australian Transport Research Forum.
2Manoj Embrandiri, Dino Isa and Roselina Arelhi, 2011. An Electric Vehicle Conversion using Batteries and
Ultracapacitors: Preliminary Performance Investigation Journal of Asian Electric Vehicles. 9(2), 1521-1527
3 Oヴデ┌┣;ヴが Mが MラヴWミラ Jが Dキ┝ラミが Jき さUノデヴ;I;ヮ;Iキデラヴ-Based Auxiliary Energy System for an Electric Vehicle:
Implementation and Evaluation," Industrial Electronics, IEEE Transactions, vol.54, no.4, pp.2147-2156, Aug. 2007
4 Rく Dラ┌ェ;ノが Lく G;ラが ;ミS “く Lキ┌が さUノデヴ;I;ヮ;Iキデラヴゲ マラSWノ ┘キデエ ;┌デラマ;デキI ラヴSWヴ ゲWノWIデキラミ ;ミS I;ヮ;Iキデ┞ ゲI;ノキミェ aラヴ
dynamic system simulatioミがざ Journal of Power Sources, vol. 126, pp. 250に257, February 2004
5Adrian Schneuwly, Bobby Maher, Juergen Auer. さUltracapacitors, the New Thinking in the Automotive Worldざく
Maxwell Technologies Inc (2004).
6Pay, S.; Baghzouz, Y.; , "Effectiveness of battery-supercapacitor combination in electric vehicles," Power Tech
Conference Proceedings, 2003 IEEE Bologna , vol.3, no., pp. 6 pp. Vol.3, 23-26 June 2003
7Dixon, J.W.; Ortuzar, M.E.; "Ultracapacitors + DC-DC converters in regenerative braking system," Aerospace and
Electronic Systems Magazine, IEEE , vol.17, no.8, pp. 16- 21, Aug 2002
8Alireza Payman, Serge Pierfederici, Farid Meibody-Tabar, Energy control of supercapacitor/fuel cell hybrid power
source, Energy Conversion and Management, Volume 49, Issue 6, June 2008, Pages 1637-1644, ISSN 0196-8904,
10.1016/j.enconman.2007.11.012.
9H.Farzanehfard and D.S Beyragh (2007), A Bidirectional soft switched ultracapacitor interface circuit for hybrid
electric vehicles. Energy Conversion Management 49(2008) 3578-3584.
10Michael George and Sam Choi (2003), Contrasting Hybrid Electric Bicycles and Electric Bicycle. Inter-Professional
Project 315 Spring 2003 p20-25.
11Golden motor (2008), Testing Report, Last retrieved:
http://www.goldenmotor.com/hubmotors/Mini%20Front%2036V200W%20Performance.pdf on 8 March 2010.
12Farnell, EPCOS - B82506W0000A007 - CHOKE, I CORE, 0.2MH, 40A, Last
retrieved:http://my.farnell.com/epcos/b82506w0000a007/choke-i-core-0-2mh-0a/dp/1644743 on 10 March 2010.
218
13Farnell, BOURNS JW MILLER - 8121-RC - INDUCTOR, COMMON MODE, 1MH, Last retrieved:
http://my.farnell.com/bourns-jw-miller/8121-rc/inductor-common-mode-1mh/dp/1692691 on 10 March 2010.
14C.Blake and C.Bull, IGBT or MOSFET: Choose Wisely. Last Retrieved: http://www.irf.com/technical-
info/whitepaper/choosewisely.pdf on 22 Sept 2009.
15W. A Giovino(2002),Microcontrollers and DSPs-Will the Two World Intersects? Last Retrieved:
http://www.microcontroller.com/Embedded.asp?did=61 on 24 Sept 2009.
16Rüdiger Bürkel and Michel Friot (1997), ASIC Based Close-Loop Transducers from 6A to 25A Nominal, Last
Retrieved: http://pccorot15.obspm.fr/COROT-ETC/Files/lts.pdf on 16 March 2010.
17 Dラミ T┌キデW がさGWデ デエW Lラ┘Sラ┘ミ ラミ Uノデヴ;I;ヮ;Iキデラヴゲざが Technology report electronic design online , November 2007.
18Croydon Cycling Campaign (2008), Electric-Assist Cycles, Last retrieved: http://www.croydon-
lcc.org.uk/info/electric.htm on 13 March 2010.